About this Author
DBL%20Hendrix%20small.png College chemistry, 1983

Derek Lowe The 2002 Model

Dbl%20new%20portrait%20B%26W.png After 10 years of blogging. . .

Derek Lowe, an Arkansan by birth, got his BA from Hendrix College and his PhD in organic chemistry from Duke before spending time in Germany on a Humboldt Fellowship on his post-doc. He's worked for several major pharmaceutical companies since 1989 on drug discovery projects against schizophrenia, Alzheimer's, diabetes, osteoporosis and other diseases. To contact Derek email him directly: Twitter: Dereklowe

Chemistry and Drug Data: Drugbank
Chempedia Lab
Synthetic Pages
Organic Chemistry Portal
Not Voodoo

Chemistry and Pharma Blogs:
Org Prep Daily
The Haystack
A New Merck, Reviewed
Liberal Arts Chemistry
Electron Pusher
All Things Metathesis
C&E News Blogs
Chemiotics II
Chemical Space
Noel O'Blog
In Vivo Blog
Terra Sigilatta
BBSRC/Douglas Kell
Realizations in Biostatistics
ChemSpider Blog
Organic Chem - Education & Industry
Pharma Strategy Blog
No Name No Slogan
Practical Fragments
The Curious Wavefunction
Natural Product Man
Fragment Literature
Chemistry World Blog
Synthetic Nature
Chemistry Blog
Synthesizing Ideas
Eye on FDA
Chemical Forums
Symyx Blog
Sceptical Chymist
Lamentations on Chemistry
Computational Organic Chemistry
Mining Drugs
Henry Rzepa

Science Blogs and News:
Bad Science
The Loom
Uncertain Principles
Fierce Biotech
Blogs for Industry
Omics! Omics!
Young Female Scientist
Notional Slurry
Nobel Intent
SciTech Daily
Science Blog
Gene Expression (I)
Gene Expression (II)
Adventures in Ethics and Science
Transterrestrial Musings
Slashdot Science
Cosmic Variance
Biology News Net

Medical Blogs
DB's Medical Rants
Science-Based Medicine
Respectful Insolence
Diabetes Mine

Economics and Business
Marginal Revolution
The Volokh Conspiracy
Knowledge Problem

Politics / Current Events
Virginia Postrel
Belmont Club
Mickey Kaus

Belles Lettres
Uncouth Reflections
Arts and Letters Daily

In the Pipeline

Category Archives

July 22, 2015

Chemical Probe Compounds: Time to Get Real

Email This Entry

Posted by Derek

There's a new paper on chemical probes out in Nature Chemical Biology, and right off, I have to disclose a conflict of interest. I'm a co-author, and I'm glad to be one. (Here's a comment at Nature News, and here's one at Science). The point of the article is that (1) many of the probe compounds that are used in the literature are inappropriate at best, and junk at worst, and (2) it's time to clean up this situation.

How bad is it? Try these examples out:

. . .For instance, LY294002 was originally described in 1994 as a selective inhibitor of PI3 kinase and remains advertised as such by nearly all vendors. Yet by 2005, it was already clear that the compound inhibited many other proteins at the concentrations used to inhibit PI3 kinase. In the meantime, a large number of more selective and more well-characterized PI3 kinase inhibitors have become available.
The availability of these new inhibitors certainly obviated the need for LY294002 as a chemical probe, and it should be discarded as a selective research tool. Yet a search
of Google Scholar in 2014–2015 alone for ‘LY294002 and PI3 kinase’ returned ~1,100 documents.

And why not? You can still find people using staurosporine as a PKC inhibitor, even though it's a kinase blunderbuss. Similarly, dorsomorphin is not a good choice to inhibit AMPK signaling, and chaetocin is a terrible excuse for a selective histone methyltransferase probe. I've written about others on this blog, as bad or worse.

But these things are all over the literature. People can't keep up, or don't, with the literature showing that these compounds (and many others) are problematic, and the suppliers keep selling them. Far too many researchers look something up in the catalog, see it listed as a "selective XYZ inhibitor", and believe every word. Both the suppliers and the investigators are at fault, and the result is that the scientific literature ends up with garbage piles and oil slicks floating all over it.

Good probe compounds are not easy to find. Seeing one in somebody else's paper places you at the mercy of their literature-searching skills if you don't do some checking of your own. Ordering one up from a catalog proves nothing more than that company's ability to sell it to you. To try to remedy this situation, this new paper also includes the launch of a web site, a wiki-based compendium of validated probes. The hope is that this will become a resource that everyone can turn to, a one-stop-shop that will save a lot of time, money, effort, and frustration.

It has only a few compounds in it as of this morning, but I plan to send in some suggestions of my own this week. (One of those is for a separate list of probes that are Not Recommended, so that people can find those as well). The plan is to put up editing functions soon so that people can do this themselves. I encourage people to send in feedback - this is an opportunity to try to fix a number of longstanding problems in the literature, and without something like this, these problems will only get worse.

Ideally, I'd like to see references to the site in the supplier catalogs, and attention paid to its listings by reviewers and authors alike. The excuses for using worthless chemical probes have never been good ones, and with any luck, there eventually won't be any such excuses left at all.

Comments (24) + TrackBacks (0) | Category: Biological News | Chemical Biology | Chemical News | Drug Assays

July 1, 2015

Lipids, Proteins, and Chapman's Homer

Email This Entry

Posted by Derek

Longtime readers might recall that every so often I hit on the topic of the "dark matter" of drug target space. We have a lot of agents that hit G-protein coupled receptor proteins, and plenty that inhibit enzymes. Those, though, are all small-molecule binding sites, optimized by evolution to hold on to molecules roughly the size that we like to make. When you start targeting other protein surfaces (protein-protein interactions) you're heading into the realm where small molecules are not the natural mediators, and things get more difficult.

But all of those are still proteins, and there are many other types of biomolecules. What about protein/nucleic acid interactions? Protein/carbohydrate interactions? Protein-lipid targets? Those are areas where we've barely even turned on the lights in drug discovery, and past them, you'd have to wonder about carbohydrate/carbohydrate systems and the like, where no proteins are involved at all. None of these are going to be straightforward, but there's a lot to be discovered.

I'm very happy to report on this new paper from the Cravatt group at Scripps, which makes a foray into just this area. A few years ago, the group reported a series of inhibitors of monoacylglycerol lipase, as part of their chemical biology efforts on characterizing hydrolases. That seems to have led to an interest in lipid interactions in general, and this latest work is the culmination (so far) of that research path. It uses classic chemical-biology probes that mimic arachidonyl lipids and several other classes (oleoyl, palmitoyl, etc.). Exposing these to cell proteomes in labeling experiments shows hundreds and hundreds of interactions taking place, the great majority of which we have had no clue about at all. The protein targets were identified by stable-isotope labeling mass spec (comparing experiments in "light" cells versus "heavy" ones carrying the labels), and over a thousand proteins were pulled in with just the two kinds of arachidonyl probes they used (with some overlap between them, but some unique proteins to each sort of probe - you have to try these kinds of things from multiple directions to make sure you're seeing as much as possible).

As well as including many proteins whose functions are unknown, these lists were substantially enriched in proteins that are already drug targets. That should be enough to make everyone in the drug discovery business take a look, but if you're looking for more, try out the next part. The team went on to do the same sort of lipid interaction profiling after treatment of the cells with a range of inhibitors for enzymes involved in such pathways, and found a whole list of cross-reacting targets for these drugs that were unknown until now.

They then turned their attention to one of the proteins that was very prominent in the arachidonyl profiling experiments, NUCB1 (function unknown, but apparently playing a major role in lipid processing and signaling). Taking the arachidonyl probe structure and modifying it to make a fluorescent ligand led to a screening method for NUCB1 inhibitors. 16,000 commercial compounds were tested, and the best hit from this led to a series of indole derivatives. These were taken back around in further labeling experiments to determine the actual site of binding on NUCB1, and they seem to have narrowed it down (as well as gotten a start on the specific binding sites of many of the other protein targets they've discovered). There are also profiles of cellular changes induced by treatment with these new NUCB1 inhibitors, along with hypotheses about just what its real function is.

Holy cow, is this ever a good paper. I've just been skimming over the details; there's a lot more to see. I strongly recommend that everyone interested in new drug targets read it closely - you can feel a whole landscape opening up in front of you (thus the title of this post). This is wonderful work, exactly the kind of thing that chemical biology is supposed to illuminate.

Comments (9) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

June 30, 2015

Two Alzheimer's Updates (Lilly and TauRx)

Email This Entry

Posted by Derek

When you look at the stock charts of the major pharma companies, there's not a lot of excitement to be had. Until you get to Eli Lilly, that is. Over the last year, the S&P 500 is up about 5%, and most of the big drug stocks are actually negative (Merck -0.4%, Sanofi down 6%, J&J down 7%, AstraZeneca down 13%). Pfizer pulled away from the index in February, and has held on to that gain (up 13% from a year ago), but Lilly - those guys were doing about as well as Pfizer until the last month or two, but have just ratcheted up since then, for a 1-year gain of over 32%. Why them?

It's all Alzheimer's speculation, as this Bloomberg piece goes into. And as has been apparent recently, Alzheimer's is getting a lot of speculation these days. BIogen really revved things up with their own early-stage data a few months back, and since then, if you're got an Alzheimer's program - apparently, any Alzheimer's program whatsoever - you're worth throwing money at. Lilly, of course, has been (to their credit) pounding away at the disease for many years now, expensively and to little avail. One of their compounds (a gamma-secretase inhibitor) actually made the condition slightly worse in the treatment group (more here), while their beta-secretase inhibitor failed in the usual way. But they've also been major players in the antibody field. Their solanezumab was not impressive in the clinic, except possibly in the subgroup of early-stage patients, and Lilly (showing a great deal of resolve, and arguably some foolhardiness) has been running another Phase III trial in that population.

They also extended the existing trial in that patient group, and are due to report data on that effort very soon - thus the run-up in the company's stock. This is going to be very interesting, for sure - it would be great for Alzheimer's patients (and for Lilly) if the results are clearly positive, but that (sad to say) is the least likely outcome. (I'm not just being gloomy for the sake of being gloomy - Alzheimer's antibodies have had a very hard time showing efficacy under any circumstances, and the all-mechanisms clinical success rate against the disease is basically zero). The same goes, of course, for the new Phase III trial itself. Things could well come out clearly negative, with the possible good results from the earlier trial evaporating the way subgroup analyses tend to when you lean on them. Or - and this is the results I fear the most - there could be wispy sorta-kinda hints of efficacy, in some people, to some degree. Pretty much like the last trial, after which Lilly began beating the PR drums to make things look not so bad.

The reason I think that this would be the worst result is that there is so much demand for something, for anything that might help in Alzheimer's that there would be a lot of pressure on the FDA to approve Lilly's drug, even if it still hasn't proven to do much. And this latest trial really is its best chance. It's in exactly the population (the only population) that showed any possible efficacy last time, so if the numbers still come out all vague and shimmery under these conditions, that's a failure, as far as I can see. No one wants to be in the position of explaining statistics and clinical trial design to a bunch of desperate families who may be convinced that a real Alzheimer's drug is being held up by a bunch of penny-pinching data-chopping bureaucrats.

And this brings us to TauRx. I still get mail about them, seven years after they made big news with a methylene-blue-based Alzheimer's therapy program. When last heard from, they were in Phase III, with some unusual funding, but there were no scientific results from them for a while. The company, though, has published several papers recently (many available on their web site), talking about their program.

Here's a paper on their Phase II results. It's a bit confusing. Their 138 mg/day dose was the most effective; the higher dose was complicated by PK problems (see below). When you look at the clinical markers, it appears that the "mild" Alzheimer's patients were hardly affected at all (although the SPECT imaging results did show a significant difference on treatment). But the "moderate" Alzheimer's treatment group showed several differences in various cognitive decline scores at the 138mg/day dose, but no difference in SPECT at all. Another paper, from JBC talks about compound activity in various cell models of tau aggregation. And this one, from JPET, is their explanation for the PK trouble. It appears that the redox state of the methylene blue core has a big effect on dosing in vivo. There are problems with dissolution, absorption (particularly in the presence of food), and uptake of the compound in the oxidized (methylene blue) state (which they abbreviate as MTC, methylthioninium chloride), but these can be circumvented with a stable dosage form of the reduced leuco compound (abbreviated as LTMX). There's apparently a ph-dependent redox step going on in gastric fluid, so things have to be formulated carefully.

One of the other things that showed up in all this work was a dose-dependent hematological effect, apparently based on methylene blue's ability to oxidize hemoglobin. It's not known (at least in these publications) whether dosing the reduced form helps out with this, but it's potentially a dose-limiting toxicity. So here's the current state of the art:

Although we have demonstrated that MTC has potential therapeutic utility at the minimum effective dose, it is clear that MTC has significant limitations relative to LMTX, which make it an inferior candidate for further clinical development. MTC is poorly tolerated in the absence of food and is subject to dose-dependent absorption interference when administered with food. Eliminating the inadvertent delayed-release property of the MTC capsules did not protect against food interference. Therefore, as found in the phase 2 study, MTC cannot be used to explore the potential benefit of higher doses of MT. Nevertheless, the delayed-release property of the MTC capsules permitted the surprising discovery that it is possible to partially dissociate the cognitive and hematologic effects of the MT moiety. Whether the use of LMTX avoids or reduces the undesirable hematologic effects remains to be determined. . .

The Phase III trials are ongoing with the reduced form, and will clearly be a real finger-crossing exercise, both for efficacy and tox. I wish TauRx luck, though, as I wish everyone in the AD field good luck. None of us, you know, are getting any younger.

Comments (16) + TrackBacks (0) | Category: Alzheimer's Disease | Clinical Trials | Drug Assays | Pharmacokinetics | Toxicology

June 26, 2015

A Look at Phenotypic Screening

Email This Entry

Posted by Derek

Here's a good overview of phenotypic screening from a group at Pfizer in Science Translational Medicine. It emphasizes, as it should, that this is very much a "measure twice, cut once" field - a bad phenotypic screen is the worst of both worlds:

The karyotype of a cell represents one of its most fundamental and defining characteristics. A large number of tumor-derived cell lines display substantial genetic abnormalities, with some extreme examples bearing in excess of 100 chromosomes as opposed to the expected 46. By that measure, the widely used human monocytic THP-1 cell line would fare well considering its overall diploid character. Nonetheless, triploidy is observed for four chromosomes and monoploidy for another, along with the entire deletion of chromosome X and substantial chromosomal rearrangements. A simple question pertains: Is this a monocyte? In other words, can we expect a faithful representation of all of the functions of a primary human monocyte from such a cell?

Using primary tissue from human patients has its own problems - availability, variation from batch to batch, limited useful lifetime in culture - but those are (in most cases) worth living with compared to the limitations of too-artificial cell lines. The authors also emphasize care in picking what ways you'll stress or stimulate the cells to mimic a disease state, and making sure that the assay readouts are as closely matched as possible to clinical end points.

The track record of gene expression readouts such as reporter gene assays is lackluster with respect to phenotypic drug discovery; no recent (>1998), first in class, small-molecule drug has originated from such an assay. A potential explanation is that mechanisms influencing gene expression represent only a fraction of all mechanisms affecting a given phenotype. . .An in-house study aimed at discovering previously unknown mechanisms leading to the up-regulation of apolipoprotein E (ApoE) secretion compared confirmed hits obtained in the same cellular system using reporter gene and enzyme-linked immunosorbent assay readouts. Although the reporter gene assay successfully identified compounds that provide large increases in ApoE secretion, it missed half of the overall hit set. . .

None of these recommendations are easy, and (from an impatient perspective) all they're doing is slowing down the implementation of your screen. Detail after detail, doubt after doubt! But your screening idea needs to be able to stand up to these, and if you just plunge ahead, you run a serious risk of generating a large amount of complicated, detailed, irrelevant data. The worse kind, in other words.

Every drug program, and every screen, rests on a scaffolding of assumptions. You'd better be clear on what they are, and be ready to justify them. In a target-directed screen, a big one is "We know that this is a key part of the disease mechanism", and (as the number of failures in Phase II show us), that's not true anywhere as often as we'd like. Phenotypic screening dodges that one, a big point in its favor, but replaces it with another big leap of faith: "We know that this assay recapitulates the human disease". You pays your money, and you takes your choice.

Comments (16) + TrackBacks (0) | Category: Drug Assays

June 24, 2015

The End of Compound Property Optimization Is At Hand

Email This Entry

Posted by Derek

Here's another Big Retrospective Review of drug pipeline attrition. This sort of effort goes back to the now-famous Rule-of-Five work, and readers will recall the Pfizer roundup of a few years back, followed by an AstraZeneca one (which didn't always recapitulate the Pfizer pfindings, either). This latest is a joint effort to look at the 2000-2010 pipeline performance of Pfizer, AstraZeneca, Lilly, and GSK all at the same time (using common physical descriptors provided to a third party, Thomson Reuters, to deal with the proprietary nature of the compounds involved). The authors explicitly state they've taken on board the criticisms of these papers that have been advanced in the past, so this one is meant to be the current state of the art in the area.

What does the state of the art have to teach us? 812 compounds are in the data set, with their properties, current status, and reasons for failure (if they have indeed failed, and believe me, those four companies did not put eight hundred compounds on the market in that ten-year period). The authors note that there still aren't enough Phase III compounds to draw as many conclusions as they'd like: 808 had a highest phase described, 422 of those were still preclinical, 231 were in Phase I, 145 in Phase II, 8 were in Phase III and 2 in Phase IV/postmarketing studies. These are, as the authors not, not quite representative figures, compared to industry-wide statistics, and reflect some compounds (including several that went to market) that the participants clearly have left out of their data sets. Considering the importance of the (relatively few) compounds in the late stages, this is enough to make a person wonder about how well conclusions from the remaining data set hold up, but at least something can be said about earlier attrition rates (where that effect is diluted).

605 of the compounds in the set were listed as terminated projects, and 40% of those were chalked up to preclinical tox problems. Second highest, at 20% was (and I quote) "rationalization of company portfolios". I divide that category, myself, into two subcategories: "We had to save money, and threw this overboard" and "We realized that we never should have been doing this at all". The two are not mutually exclusive. As the paper puts it:

. . .these results imply that substantial resources are invested in research and development across the industry into compounds that are ultimately simply not desired or cannot be progressed for other reasons (for example, agreed divestiture as part of a merger or acquisition). In addition, these results suggest that frequent strategy changes are a significant contributor to lack of research and development success.

You think? Maybe putting some numbers on this will hammer the point home to some of the remaining people who need to understand it. One can always hope. At any rate, when you analyze the compounds by their physiochemical properties, you find that pretty much all of them are within the accepted ranges. In other words, the lessons of all those earlier papers have been taken on board (and in many cases, were part of med-chem practice even before all the publications). It's very hard to draw any conclusions about progression versus physical properties from this data set, because the physical properties just don't very all that much. The authors make a try at it, but admit that the error bars overlap, which means that I'm not even going to bother.

What if you take the set of compounds that were explicitly marked down as failing due to tox, and compare those to the others? No differences in molecular weight, no differences in cLogP, no differences in cLogD, and no differences in polar surface area. I mean no differences, really - it's just solid overlap across the board. The authors are clearly uncomfortable with that conclusion, saying that ". . .these results appear inconsistent with previous publications linking these parameters with promiscuity and with in vivo toxicological outcomes. . .", but I wonder if that's because those previous publications were wrong. (And I note that one such previous publication has already come to conclusions like these). Looking at compounds that failed in Phase I due to explicit PK reasons showed no differences at all in these parameters. Comparing compounds that made it only to Phase I (and failed for any reason) versus the ones that made it to Phase II or beyond showed, just barely, a significant effect for cLogP, but no significant effect for cLogD, molecular weight, or PSA. And even that needs to be interpreted with caution:

. . .it is not sufficiently discriminatory to suggest that further control of lipophilicity would have a significant impact on success. Examination of how the probabilities of observing clinical safety failures change with calculated logP and calculated logD7.4 by logistic regression showed that there is no useful difference over the relevant ranges. . .

So, folks, if your compounds most fit within the envelope to start with (as these 812 did), you're not doing yourself any good by tweaking physiochemical parameters any more. To me, it looks like the gains from that approach were realized early on, by trimming the fringe compounds in each category, and there's not much left to be done. Those PowerPoint slides you have for the ongoing project, showing that you've moved a bit closer to the accepted middle ground of parameter space, and are therefore making progress? Waste of time. I mean that literally - a waste of time and effort, because the evidence is now in that things just don't work that way. I'll let the authors sum that up in their own words:

It was hoped that this substantially larger and more diverse data set (compared with previous studies of this type) could be used to identify meaningful correlations between physicochemical properties and compound attrition, particularly toxicity-based attrition. . .However, beyond reinforcing the already established general trends concerning factors such as lipophilicity (and that none too strongly - DBL), this did not prove generally to be the case.

Nope, as the data set gets larger and better curated, these conclusions start to disappear. That, to be sure, is (as mentioned above) partly because the more recent data sets tend to be made up of compounds that are already mostly within accepted ranges for these things, but we didn't need umpteen years of upheaval to tell us that making compounds that weight 910 with logP values of 8 are less likely to be successful. Did we? Too many organizations made the understandable human mistake of thinking that changing drug candidate properties was some sort of sliding scale, that the more you moved toward the good parts, the better things got. Not so.

What comes out of this paper, then, is a realization that watching cLogP and PSA values can only take you so far, and that we've already squeezed everything out of such simple approaches that can be squeezed. Toxicology and pharmacokinetics are complex fields, and aren't going to roll over so easily. It's time for something new.

Comments (43) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History | Pharmacokinetics | Toxicology

June 17, 2015

Organ Models on Chips

Email This Entry

Posted by Derek

Why do we test new drug candidates on animals? The simple answer is that there's nothing else like an animal. There are clearly chemical and biological features of living systems that we don't yet understand, or even realize exist - the discovery of things like siRNAs is enough proof of that. So you're not going to be able to build anything from first principles; there isn't enough information. Your only hope is to put together something that matches the real thing as closely as possible, using original cells and tissues as much as possible.

The easiest way to do that, by far, is to just give your compounds to a real animal and see what happens. But you have to think carefully. Mice aren't humans, and neither are dogs (and nor are dogs mice, for that matter). Every species is different, sometimes in ways that make little difference, and sometimes in ways that can mean life or death. Animal testing is the only way to access the complexity of a living system, and the advantages of that outweigh the difficulties of figuring out the all differences when moving on to humans. But those difficulties are very real nonetheless. (One way around this would be to make animals with as many humanized tissues and systems as possible, although that's not going to make anyone any happier about testing drugs on them. The other way is try to recapitulate a living system in vitro.

But the cells in a living organ are different than the cells in a culture dish, both in ways that we understand and in ways that we don't. The architecture and systematic nature of a living organ (a pancreas, a liver) is very complex, and subject to constant regulation and change by still other systems, so taking one type of cell and growing it up in a roller bottle (or whatever) is just not going to recapitulate that. Liver cells, for example, will still do some liver-y things in culture. But not all of the things, and not all of them in the same way. And the longer they're grown in culture, the further they can diverge from their roots.

There has been a huge amount of work over the years trying to improve this situation. Growing cells in a more three-dimensional culture style is one technique, although (since we don't make blood vessels in culture tubes) there's only so far you can take that. Co-cultures, where you try to recreate the various populations of cell types in the original organ, are another. But those are tricky, too, because all the types of cell can change their behaviors in different ways under lab conditions, and their interactions can diverge as well. Every organ in a living creature is a mixture of different sorts of cells, not all of whose functions are understood by a long shot.

Ideally, you'd want to have many different such systems, and give them a chance to communicate with each other. After all, the liver (for example) is getting hit with the contents of the hepatic portal vein, full of what's been absorbed from the small intestine, and is also constantly being bathed with the blood supply from the rest of the body, whose contents are being altered by the needs of the muscles and other organs. And it's getting nerve signals from the brain along with hormonal signals from the gut and elsewhere, with all these things being balanced off against each other all the time. If you're trying to recreate a liver in a dish, you're going to have to recreate these things, or (more likely) realize that you have to fall short in some areas, and figure out what differences those shortfalls make.

The latest issue of The Economist has a look at the progress being made in these areas. The idea is to use the smallest cohorts of cells possible (these being obtained from primary human tissue), with microfluidic channels to mimic blood flow. (Here's a review from last year in Nature Biotechnology). It's definitely going to take years before these techniques are ready for the world, so when you see headlines about how University of X has made a real, working "(Organ Y) On a Chip!", you should adjust your expectations accordingly. (For one thing, no one's trying to build, say, an actual working liver just yet. These studies are all aimed at useful models, not working organs). There's a lot that has to be figured out. The materials from which you make these things, the sizes and shapes of the channels and cavities, the substitute for blood (and its flow), what nutrients, hormones, growth factors, etc. you have in the mix (and how much, and when) - there are a thousand variables to be tinkered with, and (unfortunately) hardly any of them will be independent ones.

But real progress has been made, and I have no doubt that it'll continue to be made. There's no reason, a priori, why the task should be impossible; it's just really hard. Worth the effort, though - what many people outside the field don't realize is how expensive and tricky running a meaningful animal study really is. Running a meaningful human study is, naturally, far more costly, but since the animal studies are the gatekeepers to those, you want them to be as information-rich, as reproducible, and as predictive as possible. Advanced in vitro techniques could help in all those areas, and (eventually) be less expensive besides.

Comments (17) + TrackBacks (0) | Category: Animal Testing | Drug Assays | Toxicology

June 1, 2015

Target (In)validation

Email This Entry

Posted by Derek

Here's a short review in ACS Medicinal Chemistry Letters on target validation. As the author (Ramzi Sweis of AbbVie) says, it's easy for medicinal chemists to lose track of that being part of their role:

Too often, medicinal chemistry teams blindly accept the biological underpinnings of a new program as sound and are unaware that their ensuing efforts at compound optimization may confirm or debunk the link between the biological target and the disease. In many instances, this event occurs very early in the discovery process, at the level of preclinical interrogation.

The number of targets that you can be sure are good ones, before anyone's actually made a drug to hit them, is not large. You rarely get signals like PCSK9, where you can point at a human population with exactly the phenotype you're trying to cause (and even then, there's always the worry that your therapy might not recapitulate that "human knockout" mutation in every respect). There are a few targets kicking around out there that have never really been "drugged" but that everyone's still pretty sure would be great things to affect (p53 and/or cMyc in cancer, for example), but the list is short. Otherwise, unless someone else has already blazed that trail, you're going to be finding out not only "Is this compound going to hit the target", but "Is this target worth hitting in the first place".

Here's some good advice:

The infrastructure behind any drug discovery program should be stress-tested at a very early stage. Fear of obtaining suboptimal results from a suboptimal compound should not be a reason to defer such system testing. It is not uncommon for medicinal chemists to challenge their primary assay and note when occasional aberrations in output are observed. This feedback to biologists is quite helpful in establishing the durability of the primary assay. Accordingly, this philosophy should extend to downstream assays as well.

That fear is a very real one. "Oh, if you try to test using just the chemical matter we have now, you'll probably kill the project for no good reason". That's not a silly argument; it really is possible to go in too early with something that's too nonselective or too non-potent. But it can also be used to delay the day of reckoning, to avoid getting an answer that no one wants to get. You have to be honest with yourself and be ready to hear the bad news, and not keep telling yourself that you're waiting for a better compound before asking the hard questions. Otherwise, you'll have a compound that's been optimized in every direction and is ready for the clinic before you know if you should have been doing all that work in the first place. Some readers may just possibly have had such an experience in their careers.

But as this paper goes on to say, even trying out your compound in vivo and seeing the effect you wanted may not be enough. That's a big step, for sure, but you'll be a lot more confident if you can hit the system with more than one chemical series. (This, in my experience, is a rare luxury). It's a rare project that won't charge ahead at this point (and a rare management team that won't complain if they don't).

The certainty of target validation/invalidation is far from absolute. It is akin to a lawyer’s task of establishing judgment beyond a reasonable doubt to a jury. A highly effective medicinal chemistry team is commonly lauded for intense focus. However, this focus should not preclude a critical assessment, from the broader perspective of drug discovery, of the compounds being used to make project go/no-go decisions. . .

If you don't have that wonderful situation of multiple chemotypes, the paper suggests, then you should make the most of what you have - take some structurally similar compounds into the in vivo model, but ones that don't have activity in the primary assay. These negative controls might surprise you. That's also good advice, but can be a tough sell with some biologists. But from what I've seen, and from a look over the history of drug research in general, I'd say that skimping on the negative controls is both one of the most common shortcuts and one of the ones you'll be most likely to wish you hadn't taken.

Human nature is not built for scientific research. We have innate curiosity, no doubt about it, and we have pattern-recognition and narrative-forming abilities (and how). But what doesn't come naturally is organized suspicion of our own convictions. That has to be learned, and relearned. In this line of work, the tuition can get expensive.

Comments (14) + TrackBacks (0) | Category: Drug Assays | Drug Development

May 21, 2015

Crappy Antibodies: Available Now, and for the Foreseeable Future

Email This Entry

Posted by Derek

I made a brief mention of this article yesterday, but I wanted to highlight it. It's a look, from Nature New, at the broader implications of the antibody problem in research. Antibodies are, of course, universal reagents in molecular biology assays. If you suddenly declared their use illegal, the field would just collapse. But we can't live with 'em, either, because a really significant percentage of the antibodies used are not as good as they should be. A really disturbing percentage of the scientific literature is (at the very least) complicated by this problem, and some of it is flat-out invalidated by it.

As has been mentioned here several times, the same goes for small-molecule chemical probes, too, and how. That problem is, in principle, a bit more solvable (and I've been hearing about some efforts to try to help solve it - more on that as it develops). Small molecules are easier to assay for purity and identity, for one thing, and compared to antibodies, there are a lot fewer of them. The article estimates that there are around 300 companies selling something like two million antibodies. Which of these do what they're advertised to do, and under what conditions, well. . .that's hard to say:

Scientists often know, anecdotally, that some antibodies in their field are problematic, but it has been difficult to gauge the size of the problem across biology as a whole. Perhaps the largest assessment comes from work published by the Human Protein Atlas, a Swedish consortium that aims to generate antibodies for every protein in the human genome. It has looked at some 20,000 commercial antibodies so far and found that less than 50% can be used effectively to look at protein distribution in preserved slices of tissue5. This has led some scientists to claim that up to half of all commercially available antibodies are unreliable. . .

. . .Abgent, an antibody company based in San Diego, California, and a subsidiary of WuXi AppTec in Shanghai, China, tested all of its antibodies about a year ago. After reviewing the results it discarded about one-third of its catalogue.

So that should give you a rough estimate, and I don't think that many experienced assay development folks will be surprised. The people that are surprised, as usual, are the ones who just order out of the catalog and believe what's on the label. As the article mentions, a lot of people shop on price and speed of delivery, which (you'll be shocked to hear) are variables that don't always correlate well with reagent quality. And there are a lot of resuppliers out there, so even if you buy half a dozen antibodies against the same protein from different outfits, you may have only bought two. Or one. Who knows? And if you use up your supply of one that's working for you and re-order, will the new batch be the same as the old one? Who knows?

There are several online resources that are trying to address this problem (they're listed in the article), but many people don't even know about them. And as long people have the attitude that one (now more cautious) scientist expressed in the piece, the crappy reagents will continue to be sold. "I wasn't trained that you had to validate antibodies;" he says, "I was just trained that you ordered them."

Comments (22) + TrackBacks (0) | Category: Biological News | Drug Assays | The Scientific Literature

May 20, 2015

A Young Blood Controversy

Email This Entry

Posted by Derek

The recent revival of interest in the way that the blood from younger animals (and people?) can improve the health of older ones came bundled with a particular protein candidate for the effect, GDF11. Several papers appeared on its effects in vivo, but there were people who found that odd, according to Nature News.

Those results quickly made GDF11 the leading explanation for the rejuvenating effects of transfusing young blood into old animals. But that idea was confusing to many because GDF11 is very similar to the protein myostatin, which prevents muscle stem cells from differentiating into mature muscle — the opposite effect to that seen by Wagers and her team.

For GDF11, “You could imagine that when it came out last year that it helped muscle, it was quite a surprise,” says David Glass, executive director of the muscle diseases group at the Novartis Institutes for Biomedical Research in Cambridge, Massachusetts. “Did we miss something?”

Now in Cell Metabolism, Glass and co-workers are reporting that GDF11 does not, in fact, have the effects ascribed to it. If they're right, this is (yet another) case of antibody trouble, because they report that the antibody used in the recent papers is not as selective as it's supposed to be. Novartis has an anti-myostatin antibody in development as a therapeutic (which explains their immediate interest here) and they're looking into whether it blocks GDF11 as well. Their paper suggests that that would be no bad thing, actually.

Amy Wagers and co-workers at Harvard, though, are apparently sticking to their guns, suggesting that there might be multiple forms of GDF11. That's possible, to be sure, but it's also (I hate to say this) the sort of rationale that one comes up with after one's antibody has been called into question. I wouldn't want to count the number of times that studies (large and small) have come undone because of antibody problems, and it's almost always because they turn out to be less targeted than thought. There are a lot of proteins out there, and a lot of related proteins to any given target. Assaying an antibody against them can be quite tedious, but if you don't, you run the risk of things suddenly getting the opposite of tedious, and not in a good way. We'll see how the dust settles on this one.

If the Glass paper is correct, though, and GDF11 is not the answer, that throws the field wide open for someone to find out what the answer is. I would guess that several groups have held back getting into this area, thinking that the big prize had already been found, but if that's not so, well. . .

Update: and as fate would have it, Nature News now has a feature on the problems of antibody reproducibility! Thanks to a post in the comments section for pointing this out.

Comments (19) + TrackBacks (0) | Category: Aging and Lifespan | Biological News | Drug Assays

May 7, 2015

Fun With Thermodynamics

Email This Entry

Posted by Derek

Here's a very good review in J. Med. Chem. on thermodynamics and drug discovery. That topic has come up at this site several times over the years, and I've been wondering if anyone's reached a consensus yet. What I take away from this new article is "Nope. Not yet." Update: see also this review, also from earlier this year.

The hope, fanned by the work of Ernesto Friere and others, has been that there would be some sort of thermodynamic signature of compound binding that would help pick better compound series to work on. That's easier said than done. For one thing, you have to be able to do good calorimetry on your binding target, which isn't always possible. But even if you have plenty of protein and have the system well worked out, there are complications, and that's what this new article goes into.

Wavefunction has a good post on those, in response to this same article. One problem has been that we medicinal chemists have tried to oversimplify things, to work down to rules of thumb, and I don't think that ligand binding is amenable to that sort of thing. (Most of the really important topics in drug discovery - think toxicology - aren't reducible in that way, either, as it turns out). There is no useful textbook titled "Thermodynamics: A Hand-Waving Approach", unfortunately, because that approach doesn't work so well.

The hand-waving rules, in this case, have been "Pick the enthalpic-driven compounds", "Displace water molecules to pick up favorable entropy", among others. But there are too many exceptions. I gave a talk on entropic factors here at my workplace a few years ago, and as I dug into the topic, I found that water molecules could (and have been) assigned to every category possible, which is what everyone who takes a close look at the field discovers. Sometimes, indeed, you can displace one, and it goes flying off into entropic heaven and gives you a better T-delta-S term. But sometimes you displace one and it ends up more ordered out there in solution than it was in the protein binding site, which is counterintuitive. And sometimes (probably more often) it's the enthalpic effects of its hydrogen bonding, on both the protein and out in the bulk phase, that really determine the energetics. Thinking that displacing a water is always a good thing is just too reductionist, because the self-evident truth about water molecules is that they are not all created equal.

This is only the beginning of the complications. As the paper illustrates, in detail, similar ligands in the same SAR series (to the eyes of a medicinal chemist) can work through very different sorts of thermodynamics. Networks of water molecules, subtle effects on protein structure and stabilization, variable hydrogen-bonding networks, entropic and enthalpic penalties or bonuses from the bulk solvent, the localized water molecules, the ligand molecule, and the protein itself are all in play. You can't generalize - that's the message.

No one particularly likes to hear this. As scientists, we want to generalize, to discover trends and rules that we can use to cover more ground more quickly. As employees in hierarchical organizations, many of us would like to be able to report that we have found such things, through our own valuable keenness, and are now exploiting them (and deserve to be gainfully employed for doing so). "We've got a handle on this" is what we'd like to tell ourselves and to tell others. But when it comes to the thermodynamics of ligand binding, no, I don't think we have a handle on it yet. There may be rules and trends that will help us out, but they're not going to be as broadly applicable and wide-ranging as we'd once hoped.

Comments (14) + TrackBacks (0) | Category: Analytical Chemistry | Chemical News | Drug Assays

April 29, 2015

More on TC-2153

Email This Entry

Posted by Derek

Yesterday's post on TC-2153 and its assay activity brought a note from Paul Lombroso at Yale, whose group is doing this work. With his permission, here's an update (slightly edited):

We have now used the drug orally in nonhuman primates with cognitive deficits: it had significant results. . .we have also given it to both schizophrenic and fragile x syndrome mouse models, with the same reversal of behavioral / cognitive deficits.

TC-2153 has proven very hard to work with in PK/ADMET studies, but hopefully this will work soon. There is no change in p-tau or beta amyloid levels [one question in the comments]: an interesting finding as it suggests that inhibiting STEP is sufficient.

GSK has selected STEP as a DPAC project with me: we will be testing lead compounds using similar in vitro and in vivo models we have in the lab. They were not interested in TC-2153, nor have I been able to get NIH funding for this compound (the structure makes medicinal chemists nauseous).

So we continue to use it in the lab as a useful tool, while searching for new STEP inhibitors.

Sounds like an interesting target, and a good field to be working in. My thoughts are that the compound itself should be run through as many assay panels as possible (commercial and in-house) to see just what other effects it may be having. If it's going to be the basis for a lot of work on STEP, it needs to be characterized as well as possible (especially with that structure and mode of action). And as mentioned yesterday, other classes of Cys-targeting covalent inhibitors might well be very useful. I'll keep an eye out for updates in this area.

Comments (16) + TrackBacks (0) | Category: Drug Assays | The Central Nervous System

April 28, 2015

Sulfur, Sulfur, Sulfur

Email This Entry

Posted by Derek

Does anyone know of any phosphatase inhibitors that aren't hideous? I ask this because someone sent along a question about this paper, from last August, that I'd missed at the time (press release here, but the paper's open-access as well). Here's a commentary in the journal itself. It's work from Yale on an enzyme called STtriatal-Enriched tyrosine Phosphatase (STEP), a brain-specific enzyme which has been thought to be involved in memory and other CNS processes. The group was looking for an inhibitor, and screening for phosphatase inhibitors has been, historically, No Fun Whatsoever.
I'd say that this is still the case, given that the best molecule that turned up was TC-2153, shown at left. I've seen several molecules in this polysulfide class over the years, but I've never seen one that went very far. It's not necessarily an impossible molecule, but it's going to need some special care in development. You'd want to look out for oxidation (both on storage and metabolic), exchange reactions with other thiols, nucleophilic ring opening, and so on. The compound appears to work, in fact, by reacting with the thiol on a catalytic Cys residue in the enzyme's active site.

This structure was arrived at by an unusual route: during the initial screen, several compounds were apparent hits. Resynthesis of them, though, gave no activity (a common experience when working with difficult targets!) Checking the original samples turned up an impurity that actually showed potent activity. This was fortunate - you often go back to these things looking for such a clue, but finding something identifiable isn't easy. Often it turns out to be some sort of polar polymeric gunk (like this), colorful stuff that sticks to silica on a quick filtration, but in this case, it was a yellow crystalline substance that eluted late. X-ray crystallography of it showed it to be elemental sulfur, S8, which must have been an interesting surprise. Its IC50 was 17 nM! Potency aside, that's really going to have trouble as a CNS-targeted drug, as the team well realized, so they deliberately had a look at the benzopentathiepins as surrogates. (As a side note, the authors point out that they didn't include DTT as a reducing agent in their assay conditions - had it been there, they very likely would never have picked up on this compound class at all). It's also worth noting that when you do find the Amazing Active Impurity, the chances are increased that it's working through a covalent mechanism.

TC-1253 itself showed good activity in cells and in mice. But it also shows evidence of off-target effects at higher doses, as well it might, and shows some activity against related protein tyrosine phosphatases in vitro. (It looks better in cells and in the mice than it does against the straight enzymes, which might be the same sort of effect that the Cravatt group saw with covalent inhibitors as well, moving from proteins to cells).

Given these results, I'd suggest going full-on covalent for the discovery of further STEP inhibitors. Since that catalytic Cys is important, and indeed seems to be more nucleophilic than usual, why not go after it with a big array of acrylates, weak leaving groups, and so on? The tolerance for such mechanisms has increased over the years in med-chem. There are a lot of interesting Cys groups out there, and sufficiently diverse compounds should be able to target them selectively. It's tricky work - the line between a lead compound and an assay-interfering compound is not a clean one - but what other success has anyone had against phosphatases, anyway?

Update: more on this compound here.

Comments (16) + TrackBacks (0) | Category: Alzheimer's Disease | Drug Assays | The Central Nervous System

April 8, 2015

Colorful Compounds Strike Again

Email This Entry

Posted by Derek

Here's a paper of the sort that I wish more people would publish: a report of a false positive compound, and why it was a false positive. The authors, from the Helmholtz Institute at Saarland University, were screening for compounds that might directly inhibit bacterial RNA polymerase. That's certainly a worthwhile target, although (like the other bacterial enzymes), it's sure not an easy one. Rifampicin is probably the prototype antibiotic with this mechanism, and not many of us have sent things into the screening collection that look like it does.

pyrrole.jpgBut in a fragment screen, the Helmholtz folks came across the small molecule shown, with pretty decent potency and ligand efficiency. Isomers of the same structure had already been reported as polymerase inhibitors, as it turned out. But when they made some more of the compound, well. . .it wasn't active.

On standing, though, the newly synthesized material began to take on some color, gradually turning redder and redder. Now, you don't get much redder than rifampicin, which can even stain your sweat, but redness is not a necessary property of a polymerase inhibitor. Suspicious, the team aged some of the material by warming it up in an open flask, and sure enough, the activity in the assay was back. And the redder the stuff got, the better it worked. Analysis showed some high-molecular-weight polymeric material forming - their proposal is that the pyrrole is polymerizing and leaving a backbone of carboxyl-substituted aryls poking off in every direction. I can well believe that. I've been on the receiving end of some pyrrole-containing false positives myself, and in several of those cases there was some red-purple stuff that, once removed, killed the activity.

This carboxylate-studded gunk apparently is a good nucleic acid mimic. Gel-shift studies showed that it binds tightly to RNA polymerase, with a Ki in the low nanomolar range. (RNApol, of course, is ready and looking for regular repeated negative charges). A quick check, though, showed that the polymer was a strong inhibitor of a bacteriophase RNA polymerase, but not those from some other species. Still, you could expect that something like this could well tangle up your assay results if your target protein has a net-positive-charge region that's important.

It's tempting to just say "Fine, no pyrroles" after seeing things like this. But it's hard to make rules like that when a pyrrole (albeit a more heavily substituted one) has gone into millions of patients (and pulled in tens of billions of dollars in revenue). Probably the best lesson to draw is to be suspicious of lightly substituted pyrroles, especially if the sample has any color whatsoever, and always be ready to resynthesize/repurify them you start high-fiving anybody.

The Helmholtz group, by the way, did go on to find some things that were more robust. But I'm glad that they published their blind alley along the way - it could do someone some good on some totally unrelated project.

Update: in response to questions in the comments, the authors did try their polymeric stuff against actual cells, but with no effect. It surely has severe problems getting across the membrane, but if you could get it in, I'll bet it would do all kinds of things. . .

Comments (22) + TrackBacks (0) | Category: Drug Assays

April 2, 2015

Silicon Valley Sunglasses

Email This Entry

Posted by Derek

The intersection between Silicon-Valley-style tech and biotech is getting a lot of attention these days. Some of it looks like it could be a productive synthesis: 23andMe hiring Richard Scheller of Genentech (update: and Robert Gentleman, today) as it starts its own efforts in drug discovery and Google bringing on Art Levinson (Genentech's former CEO) to run its Calico subsidiary. (They recently signed a big deal with the Broad Institute and one with the University of California system). You'll have noticed that I'm defining "productive" as "listening to people who've done it before", and I make no apologies for that.

That's because, as I've gone on about before, drug discovery is a very different world than hardware and software. People working in those areas have gotten used to the key capabilities of their work improving wildly and constantly getting cheaper, tendencies that have rippled out to affect everything than has a data-crunching rate-limiting step in it. And it turns out that all kinds of things do (or can be reconfigured to). But living in this environment can give a person an unusual perspective on the world.

Some of that perspective is welcome - the idea that there are always huge opportunities out there waiting for someone with enough speed and nerve to go after them, for one. That's very Silicon Valley (and it's also very American) and I think it's great. But if along the way you pick up the idea that the world of apps, code, and processor speed is the default setting for the world, you can start to see everything that doesn't advance that way as defective. That's the Andy Grove fallacy, as I've called it (referenced in those links above), the idea that understanding human disease and its treatments should be pretty much like designing a new chip or writing a new app.

There's another problem that's not unique to the Valley, although it does tend to give people a bad case of it. That's the "Clearly I'm smart and successful, so clearly I have something to offer in this other field over here" one. We all succumb to that one now and then; it's human nature. You can watch Mark Cuban display it here, with respect to medical testing.

But here are a couple of recent examples of the more localized problem. I wrote last year about Emerald Therapeutics, an outsourced-lab-assay company backed by Peter Thiel (who may also be interested in their antiviral therapy ideas). Here's another article on them, and it asks, in so many words, "Why is new drug development so comparatively torpid when app development is so torrid?". I couldn't provide a more succinct version of the Silicon Valley/biopharma disconnect if I tried.

According the article, the folks at Emerald ". . .think it comes down to the difficulty of running experiments in the life sciences". But I'd like to propose that this difficulty, at least for early-stage work like Emerald is proposing to do for people, is largely a matter of contrast. If you're used to being able to sit down and bang out code, any time, anywhere, with all kinds of tools (libraries, compilers, virtual machines, what have you) at your fingertips, then yeah, working up a new assay protocol in a cell line is going to seem agonizingly slow. Multibillion dollar ideas can be cranked out in the coding world very quickly, if you hit the right place at the right time, but just you try that in the lab. Now, I have no problem with Emerald running assays for people, although it may yet be harder than they're thinking. But they're not removing as much of a bottleneck as they might think. The real bottlenecks are figuring out what assay to run, and what to do with the data once you have it. Can't outsource those.

Then we have this piece on genomics at TechCrunch. Experienced readers will get a very 1999 vibe from it; it's full of the wonders-of-the-genome stuff that was so thick on the ground back then. (That makes it simultaneously cute and annoying to see it all again, of course). "One App Away" is a headline from the article, which is suffused with the next-big-thing attitude that you'd find from someone who's figured out how to social-gameify the process of splitting the check at a bar, complete with a three-second clip of everyone hoisting drinks with the amount they chipped in hovering over their heads. That probably already exists.

But you know, I don't really disagree that much with many of the conclusions in the genomics piece - just the pace at which things will happen. That's what I think has been Valley-ized there, the idea that very, very soon now something will just wildly, exponentially take off. As much as I might like to see something like that happening in biopharma, though, I can't quite make myself believe it. Technology, Silicon Valley style technology, is human-designed and human-optimized for other humans. As human beings, we're playing on our home turf there. But the biology of disease is an away game if there ever was one. The inner workings of cells and the ways that they work together are flat-out alien compared to anything we've ever built ourselves. People who are used to coding up apps have never experienced anything like it, and many of them don't seem to realize that they haven't. Expecting the sorts of behavior that you get from human-built technologies, and expecting the same effects from the techniques that work to optimize them, is an expensive accident waiting to happen.

Comments (35) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

April 1, 2015

Rotten Rottlerin

Email This Entry

Posted by Derek

Here's one of the problems with the various schemes to do "deep learning" using the published scientific literature: a fair proportion of that literature is junk. This is not news, of course, and to some extent I feel that that's always been the case. The pressures to publish, and the corresponding rise of publishing fraud, aren't helping much, though. And there's a lot of plain old sloppiness out there, too.

For a small example, consider the polyphenol compound rottlerin. It's isolated from a plant in India and the surrounding area (and named after the botanist who first described the genus in the European literature). Rottlerin itself was isolated in the 1800s as a tan, crystalline substance, and in the early 1990s it was identified as a PKC-delta inhibitor, for which purpose it has been used in uncounted studies in the literature.

Problem is, though, like a lot of such compounds, it's not anywhere near as selective as advertised. And it has all sorts of other activities in cells, too, as one can well imagine, looking at its structure. In fact, as that link shows, the original PKC activity reported for the compound may well have been due to an impurity - other papers have found no real PKC inhibition at all, much less anything selective (not that the original selectivity was anything to jump around about, either). So rottlerin, as a tool compound, is useless. It does too many thing to untangle, and the the thing that it's most well-known for doing it probably doesn't do at all.

So no one uses it any more, right? Har de har. That link will take you to the latest list of rottlerin papers, and on it you'll find this paper from this year, using it as a PKC-delta control. And this 2015 paper. And this one. And this one. You can pick out more of them, all the way down the list.

Now, some of the papers you pull up are more defensible, since they're more like "Here's something rottlerin does phenotypically to cells, and we're going to unravel what it is". Fair enough. But why is anyone using it as a PKC-delta inhibitor, a good dozen years (at least) since that activity was called into question? I wouldn't trust it as a pan-PKC inhibitor either, not when it does so many other things. Rottlerin is useful if you're studying rottlerin, but getting it to illuminate something else for you is going to be a tricky business. Watching people just blithely treat it as if it were some automatic selective switch you can drop into an assay is disturbing.

That broad phenomenon has come up in discussions around here before, though. There's a persistent lack of examination in the biological world when it comes to small-molecule tool compounds. Too many people seem willing to believe that there are all sorts of those perfectly selective compounds out there, ready to reach into cells and do exactly what it says in some catalog or some old paper. There are some, but nowhere near as many as you might imagine, or as many as people will sell you. Tocris is more up front about what the compound does, but here's EMD/Millipore/Calbiochem, selling rottlerin as " cell-permeable and reversible protein kinase C inhibitor that exhibits greater selectivity for PKC-delta (IC₅₀ = 3-6 µM) and PKC-theta". So how can I blame some biologists for not trusting that sort of thing?

Update: from the comments, here's a complaint about the use of this very compound from 2007. It spells it out in the title ("An Inappropriate and Ineffective Inhibitor"), and it appeared in what a colleague of mine once called "An obscure journal called Cell"). But rotterlin itself marches on, and neither editors nor reviewers seem to care, in many cases.

Comments (27) + TrackBacks (0) | Category: Biological News | Chemical News | Drug Assays

March 19, 2015

Ignoring the "No Loitering" Sign

Email This Entry

Posted by Derek

How long does a drug molecule stay in its site of action? How long does it need to? These questions have had a higher profile in recent years, as more drug discovery teams pay attention to the idea of residence time. This is a short review, with an impressive list of examples from the recent literature, and it's a good place to get caught up on the topic.

You'd think that residence time would be closely related to the binding constants that we already measure - high potency would equal long residence time - but it's not always that straightforward. Even rather similar structures (with rather similar potencies) can have significant differences in off-rates and residence times, for reasons that are not always easy to work out unless you have a lot of high-quality structural and kinetics data (and sometimes not even then). One reason is that a compound binding to a target is not always the simple binary complex that we tend to picture. You can have a molecule that comes in and binds to one conformation of the protein, only for that event to set off a shift to another conformation that can bind the drug more tightly or more loosely (and through different interactions). Or perhaps the binding site is in equilibrium between several low-energy states even before the drug molecule shows up, but some of these states are competent to bind it, and some aren't. The ways in which a ligand can perturb these balances can be subtle, but important.

And then you have the princes and the kings of low-off-rate compounds, the reversible covalent compounds and the irreversible binders. These have distinct profiles in careful kinetics experiments, but (as this review mentions) there have still been many reversible binders that were not recognized as such until comparatively late in the game. Both types of compounds have moved from curiosities, to be treated with some wariness, to perfectly acceptable (and even sought-after) classes over the years.

But don't think that this is an guarantee of success, because the one constant in drug research (from Paul Ehrlich on up) is that there is no such thing. As this review shows, there are also a few examples of compounds whose in vitro behavior showed clear and reproducible variations in residence time, but whose behavior in animal models (or humans) hardly budged. Among other things, this effect can depend on the tone of the system you're trying to regulate: does it need to be inhibited for a long time, or will a quick mash on the pedal accomplish the same thing? Needless to say, this is often a question that we don't know the answer to up front - the only way to find out is to make some compounds and see what they do.

Which is, in the end, about the size of it for drug research in general. It would be nicer (and certainly cheaper) if we didn't always just have to find some compounds and see what they do, but often that's just what it comes down to. For many targets, though, you'd probably do well to consider residence times as one more way to get where you're going.

Comments (15) + TrackBacks (0) | Category: Drug Assays | Drug Development

March 18, 2015

Singlet Oxygen PAINS

Email This Entry

Posted by Derek

Here's Peter Kenny on an important aspect of the concept of "PAINS" in screening (the sort of pan-assay-interference compounds that have been a big topic in the field the last few years). He notes that the original PAINS paper grew out of some campaigns using AlphaScreen technology. And although some of the classes of compounds that have been flagged probably hit other assay formats as well, to their disadvantage, some of them may be peculiar to that platform. Especially worth noting are compounds that can scavenge singlet oxygen, because the AlphaScreen mechanism will cause those to ping up as false positives. The various double-bond-to-sulfur compounds that many of us love to hate may well be acting through that mechanism.

He's certainly right about this, and his recommendation for AlphaScreen users to read up on the single oxygen reactivity literature is worth seconding. I think, though, that there are more reasons to dislike those sorts of sulfur compounds than just this - they have a bewildering set of reactivities and instabilities under various assay conditions, which is really what makes them so pernicious as screening hits. In fact, that's one the key concepts that those of us who are big on the PAINS concept are hoping to communicate to other chemists, to biologists, and others: that (1) there sure are a lot of ways that you can get false positives, (2) there are compounds that are capable of performing more than one of these tricks simultaneously, so (3) these things tend to over-represent as plausible screening hits, but are difficult-to-impossible to carry on with.

Comments (2) + TrackBacks (0) | Category: Drug Assays

March 10, 2015

A National Compound Collection

Email This Entry

Posted by Derek

I remember when I was in my grad school research group, looking up at the shelves of lab notebooks from past students and postdocs. Some of these were taken down and used from time to time - "Oh yeah, so-and-so made that intermediate one time, must be in one of his notebooks" - but many of them rested undisturbed. The same goes for many a PhD dissertation (I know mine hasn't troubled very many people, that's for sure).

At the time (and we're talking mid-1980s, dang it all) I kept wondering if there were some way to get at all this information and make it searchable, especially by structure. Since my group worked in a fairly specialized area of organic synthesis, many of the former reactions had at least a chance of being relevant to some future grad student. WIth 1985 hardware and software that was a bit of a tall order, but when I use my current electronic notebook I'm seeing those thoughts made real, since I can search company-wide by structure (or whatever other criterion I can dream up).

But there are certainly masses of unpublished data sitting out there, and a group in the UK is trying to bring it out into the light. In a pilot project, they've gone over 750 PhD dissertations from 15 universities, and extracted 45,000 structures (75,000 if you count all the stereoisomers). This effort is funded by the Royal Society of Chemistry, so the structures are going into ChemSpider (and most of them were new to it, as well they might be).

As the article details, the original hope was for a physical collection of compound samples, but a variety of issues - not least, cost - has made that hard to realize. So this is a virtual set so far, which would be available for virtual screening. And here's where opinions really start to diverge. 75,000 physical compounds makes for a pretty good storage and dispensing effort, but that many virtual compounds is a tiny drop in a very large bucket. And if you'd like to double, triple, or quadruple that number of compounds, it would be short work, computationally. Of making virtual screening sets there is no practical end. (One wonders how many of the RSC dissertation compounds are already in the GDB sets). It's the chemical space problem - if you're just filling it out randomly, there's an awful lot to fill.

Now, that same argument can be applied to a physical compound collection, and such a collection will inevitably be even smaller than a virtual set. But it has a big advantage in utility, because the fact is that we really can't do virtual screening with a great deal of confidence yet. Real compounds against real proteins, that's the way to go if you can do it. The expense will be far greater, though. Just collecting the compounds themselves will be a major effort, and you need a building to store them and all the equipment that big screening collections call for. The curation of the set will be an even bigger pain: it's absolutely certain that a reasonable percentage of such compounds will be clean, but not what they say on the bottle, and what may be an even larger percentage will not be very clean at all. How to deal with those?

From what I can see, the RSC is working on those questions, and I wish them luck (and funding). What if everyone who did work in a publicly funded lab in the UK sent samples in to the National Compound Collection? What if we did that in the US? We have a lot of chemistry going on in academia, and untold numbers of compounds that get squirreled away in vials and stuck in desk drawers. How much money would it take to get them brought together, and would such an effort pay off?

Comments (21) + TrackBacks (0) | Category: Chemical News | Drug Assays

March 2, 2015

Of Proteasome Inhibitors and PAINs

Email This Entry

Posted by Derek

Amgen is out with some new data that might well justify their purchase of Onyx a year and a half ago. A big driver for that deal was the proteasome inhibitor Kyprolis (carfilzomib), and the company just reported results in a head-to-head trial in multiple myeloma versus the Takeda/Millennium competition, Velcade.

Kyprolis comes out looking pretty good:

In the "ENDEAVOR" study announced Sunday night, patients with multiple myeloma still progressing despite one to three prior therapies were randomized to receive Kyprolis plus a steroid or Velcade plus the same steroid. Following an interim analysis, patients in the Kyprolis arm had a 47% reduction in the risk of disease worsening or death compared to Velcade. At the median, Kyprolis patients went 18.7 months before their disease progressed compared to 9.4 months for Velcade patients. The benefit favoring Kyprolis was statistically significant.

The study is continuing to see what the overall survival benefit might be, but I'm sure that Amgen is hopeful that those numbers will translate into something robust. Velcade itself will be off patent in a couple of years, so Amgen is going to need data to make the case that people should get their drug rather than the generic competition.

It's worth taking a look at the structures of those two drugs again. Velcade is, of course, a boronic acid, the first to get approved as a human therapy. "Boron is for morons" went the joke for many years, as boron-containing enzyme inhibitors ran into trouble on their way to the clinic. But that particular moronic compound brought in over two billion dollars in sales last year, which a lot of us smart people who've avoided boron have never been able to ring out with anything we've made. And Kyprolis is rather funny-looking, too. It came out of some natural products work in Craig Crews' lab at Yale, and it's a modified tetrapeptide with an epoxide hanging off it. This is another structure that would get the fisheye from a lot of people, for both those reasons, but there it is, out there in the clinic working well.

Now, I take the point that targeting the proteasome is not exactly like coming up with a new diabetes drug. You're going to be treating some very sick patients, many of whom are (otherwise) going to die quickly. The sorts of structures that a project is willing to look at do need to be calibrated a bit for these things. But a lot of us - including me, a few years ago - would have calibrated these two drug structures right off the side of the page, and that clearly would have been a bad decision. Yet another reminded for us to loosen up a bit.

"Right", I can hear some readers saying. "Here's one of those guys who's death on PAINs and makes fun of people's screening compounds, telling us to loosen up on funny structures". A fair point, but here's where I get off saying this. The big rap on boronic acids and peptidic drugs is that they have poor PK, and that's something that can be checked out. You'll note that morpholine hanging off the end of carfilzomib, and I suspect that's on there for just those PK reasons. Epoxides, for their part, are actually a lot less reactive and nonspecific than their reputation has it.

PAINs, though, are not looked at with distaste becuase of their phamacokinetics or how they've tended to perform in the clinic. They're trouble because they tend to give false screening results and/or hit in way too many assays to be good candidates for further development. So when they show up in a paper as great new screening hits, and the authors show no sign of realizing their problematic nature in just the sorts of assays that they've been running, then yes, it's a problem. Anyone developing a boronic acid, a tetrapeptide, or an epoxide should know that they have a lot of PK and tox assays waiting for them, and that no one will believe in these compounds until they start passing them. For PAINs, this disbelief kicks in very early, as is should. The very first step in the whole process, activity in the screening assay, may well be bogus.

And in the same way that there are PK and tox assays to sort out odd-looking structures that have been known to have trouble in these areas, there are screening-level assays that need to be run for suspicious-looking hits. They need to be shown not to be redox cyclers, not to hit all sorts of other targets, not to decompose to reactive species under the assay conditions, not to soak up thiol nucleophiles nonspecifically, not to hit because of fluorescent interference, and so on, and so on. Peptides and boronic acids have a reputation for not clearing the later hurdles in drug development, so if you're working on them, be prepared and let the data guide you. They'll work, or they won't, and you'll get a clear answer. False positives are not a big problem in most PK assays.

But PAINs are the compounds that tend to not even clear the first hurdles, while looking like they have. Let the data guide you there, too, but that means getting the correct data with those follow-up assays after your initial screen, not declaring them wonderful new leads and pushing on regardless.

Comments (6) + TrackBacks (0) | Category: Cancer | Clinical Trials | Drug Assays | Drug Development

February 27, 2015

PAINs And Good Old Med Chem

Email This Entry

Posted by Derek

The mainstreaming of the PAINs concept continues, with editorials from Jonathan Baell in ACS Med. Chem. Letters and Dan Erlanson in J. Med. Chem.. Both are definitely worth a read.

Baell emphasizes that real hits tend to have real SAR around them. You can take pieces off the structures, and activity falls away; you can move things around and it changes. Eventually, you can make more potent compounds. I will say that I've seen some oddball programs where things resolutely refused to make sense, and the projects were advanced by a sort of human-wave attack. But even in those cases, the compounds were, in fact, advanced, even it it wasn't always sensible how they got there. PAINs compounds, on the other hand, just seem to sort of sit there. You try this, you try that, you try the other damn thing, and you never seem to get much better than the original hit, even though the activity might wander around some with structural variations.

Their hallmarks typically comprise the following: little or no medicinal chemistry optimization, unconvincing structure–activity relationships (SAR), relative lack of improvement in biological activity to meaningful levels that often hover around the micromolar mark, and molecular modeling described as though it is an experimental observation of relevant binding sites. Also, the literature is frequently ignored as an important SAR source of evidence that similar compounds appear to be hitting different targets and could be promiscuous.

Baell also makes a good point about the size of many screening efforts, particularly academic ones. It's a worthy cause to screen against a tough target, but you should be prepared to go big or go home:

Screening too few compounds, perhaps even as small a number as several thousand, is a contributory problem. An understandable attraction to academic researcher with a tight budget is the relative affordability and that such an exercise may not even require access to robotics but just the services of an on-site research assistant. However, screening too small a library of unbiased commercially available compounds may not return a progressable hit. In contrast, it will certainly produce artifacts that are then more likely to attract unwarranted further attention. . .

. . .I believe that even screens of 30,000 compounds are suboptimal, especially for difficult targets. For highly druggable targets, some progressable hits will plausibly be unearthed, but why not screen 150,000–250,000 compounds to find better hits that could potentially save years of medicinal chemistry optimization?

Well, money, for one thing, but he's right that most of the expense of doing a 250K screen will have already been encountered while doing a 30K one. And if your assay has false-positive problems, running it across a large collection will merely bury you in more junk, which is all the more reason to take care before you start screening at all. The point stands, though: if you want to go after a protein-protein interaction or a transcription factor (or some such), a 10,000 compound screen is completely inadequate. Update: as has been pointed out in the comments, sheer numbers alone are not necessarily enough. It is possible to have large piles of horrible compounds, and making them larger is not the way to go. And there is no way to run a "focused library" screen to improve your chances (no matter what the vendors might tell you), because no one has any good idea of what to focus on for such targets. Going on to cell-based and phenotypic screens can actually exacerbate the problem, because it opens up new ways to get fooled (membrane-perturbing amphipathic compounds, etc.).

Erlanson's piece is a commentary on the recent paper discussing thiol-scavenging compounds as promiscuous screening hits. He makes the point that what looks like SAR may just be SIR (a structure-interference relationship), since there are no classes of PAINs where every member hits the assay. You have to run counterscreens, control experiments, and mechanistic tests, particularly if you're looking at a compound that might be reactive (like a thiol scavenger). And even experienced chemists might not realize that they're looking at one, so run those experiments anyway, for peace of mind and to avoid wasting even more time down the line. Reactive compounds can indeed be great drugs (look at ibrutinib, or heck, look at aspirin), but they can also spit out confusingly convincing false positive data in a screen and scuttle your chances of finding anything real.

Another point that Erlanson's article emphasizes is that computational screens of structure, while useful, are always going to be inadequate. It's hard to write such filters so that they catch everything that they're supposed to catch, for one thing, but the bigger problem is that there are many more problematic structures than are contained in any set of filters at all. There is, alas, no substitute for doing more work and taking more care, and this annoying truth has been with us for a long, long time.

Baell makes a similar point, that structural filters are necessary but nowhere near sufficient. And I particularly liked this part:

We need to recognize the astonishingly limiting bottleneck created by the pace with which hit discovery through HTS so overwhelms the subsequent and necessary medicinal chemistry optimization. Impatient biologists may find it hard to understand that HTS heralds just the start of a multiyear journey of medicinal chemistry optimization. In this context, the medicinal chemists among us could do a better job explaining to biologists that drugs are not discovered through screening or design, but principally through medicinal chemistry.

Preach it! And let this message not only be heard in the biology labs, but yea, even unto the upper reaches of management. None of our spiffy equipment, none of our advanced techniques will do the job by themselves. Someone has to make compounds, and then make some more, and there's no way around it. A significant part of the history of the last 25 years of the drug industry consists of attempts to evade this part of the process, but they haven't worked, and they're not going to.

Comments (30) + TrackBacks (0) | Category: Chemical News | Drug Assays

February 16, 2015

Targeting a Transcription Factor

Email This Entry

Posted by Derek

Here's a paper in Science on inhibition of the transcription factor CBFbeta-SMMHC. That's a messed-up version of CBFbeta itself, which has been found to drive many forms of acute myeloid leukemia (AML) though its interactions with another transcription factor, RUNX1. So tying this one up with some sort of small-molecule therapy is an attractive idea, the only problem being that getting small molecules to work on transcription factor pathways has been very difficult indeed.

Transcription.jpg This group set up a FRET assay and screened the National Cancer Institute's diversity set of compounds to see what they could come up with, and 2-(2-pyridyl)-5-methoxybenzimidazole same out as a micromolar hit. (Without the methoxy, it lost at least 10x potency). Since the transcription factor itself is dimeric, they then tried stitching two of these together, with the results shown. All of these are in the 300/400 nM range. There's a good deal of downstream biological data suggesting that these compounds are hitting this pathway.

So, what does a medicinal chemist think about these compounds? For starters, 2-pyridylbenzimidazole is going to be a very good metal chelator, and I would wonder if some of the effects might not be due to an off-target mechanism of that sort. The paper itself does not mention any possibility of metal chelation, as far as I can see. The 5-methoxybenzimidazole itself isn't objectionable per se (that's the left-hand side of omeprazole), but I have seen assay hits with that sort of chelating group in them, and in my own experience they've been difficult to prosecute. (I'd be glad to hear some other opinions).

Now, as far as the dimer idea, since the target itself is a dimer, it's hard to object to the idea. But for whatever reason, this strategy doesn't seem to have paid off a lot in drug discovery. Perhaps it's the added molecular weight and PK properties that you pick up when you do this that make them harder to advance. On top of this, medicinal chemists have a sort of instinctive distaste for completely symmetric molecules like these - and I do, too - but I've never been quite sure why. I think that we feel that there are more reasons for something like this not to be real than for it to be acting as advertised, but that's not a very quantitative look at things.

I would feel better about this latest paper if the chemical matter had been given more of a going-over than appears to have been done. I don't see any mention of selectivity assays or the like, and I really would like to run something like this across some sort of broad panel just to make sure that it doesn't set off too many other things. On the other hand, though, the authors do have a lot of cell data that points toward this particular transcription factor being the target, and I'm really to applaud anyone who gets the small-molecule transcription factor thing to work. I'll look forward to seeing what comes of this, for sure.

Comments (21) + TrackBacks (0) | Category: Cancer | Drug Assays

February 6, 2015

A New Dual DNA-Encoded Library Screen

Email This Entry

Posted by Derek

Neri%20scheme.jpgI've written several times about DNA-encoded chemical libraries. Now comes an interesting new variation on them from Dario Neri and Jörg Scheurmann of the ETH in Zürich, pioneers in this area.

As you can see from that scheme, what they're doing is trying for secondary binding sites. For many protein targets, the ones without evolutionarily honed small-molecule binding regions, that may well be the only way to get enough potency to be useful. That's the idea behind fragment linking (as in the early days of "SAR by NMR" (see this PDF for some references), and several other schemes as well). It's always been an appealing idea, because mathematically, if you get the full binding of each component, linking them gives you the product of the potencies. So a 100 micromolar compound, linked perfectly to another 100 micromolar compound, could give you a ten nanomolar hybrid, which is an attention-getting prospect.

In reality, it's rare that you can get this to work, though. Getting two separate ligands to bind simultaneously is no small feat, to start with, especially if you run them one at a time. The binding of compound A may well affect the shape (or even the existence) of the binding site for compound B, so one-step-at-a-time assays don't always line up. If you're testing mixtures, though, it can be hard to figure out when two things are binding simultaneously, depending on your assay method. And even when you can, how do you determine where those two sites are, how close they are, what the orientation is of the two ligands, and how best to connect them? That's where many of these ideas have run into trouble.

One way around this is dynamic combinatorial chemistry, DCC, where you have a big mixture of components that can form reversible bonds with each other. (The two systems used the most have been thiols/disulfides and oxime formation). That gets around the orientation problem, since the only ones that form to a greater extent in the mixture are the ones that can form via cooperative binding. The problem that arises then, unfortunately, is that disulfides and oximes are not very good tool compound or drug candidates. There's the reversible decomposition, for one thing. And trying to replace those linkers is no small feat. There's nothing that looks and acts quite like a disulfide. You'd think that running three methylenes or so in there instead of a single S-S bond would do the trick, but it mostly doesn't. And even modifying your starting fragments to take that sort of linker isn't always straightforward.

A technique to get past that problem is the Barry Sharpless in situ click chemistry, which I've always had a high opinion of. In that system, you let a host of acetylene and azide-containing fragments loose against the protein target. The triazole formation is so slow at room temperature that you have a very low background rate of reaction, but if two components can bind in a way that their reactive ends get jammed together in the binding site, you can see enhanced amounts of the product by LC/MS. This way, you still don't have to know the orientations and spacing of the two components, and you form a linkage between them that's pharmacologically acceptable. This has worked in quite a few examples, sometimes quite dramatically, but it certainly doesn't always work, either.

In this latest work, as you can see, the two library components are each on their own strand of DNA. That's been tried before (with Neri's ESAC system) but it's often been tricky to read out the result from a large library. The way this system is set up, though, the A strand has its "bar code" in a region where the B strand has a linkage that is incompatible with any later PCR. Then, further down the strand, the B code is set up so that the corresponding A strand can have that B code copied over to it by a Klenow fill-in reaction. That's a neat trick. What it gives you, then, is a system where if you get good binding, with both components contributing, you can wash off all the other things that didn't work, run the Klenow, and then read off the A strand by PCR, which now simultaneously tells you the identity of the A and B components via their corresponding DNA coding.

The A library in this paper has 550 pharmacophores, and the B has 202, so that gives you about 111,000 combinations. (And the fact that you're relying on DNA hybridization to drive things means that you get much closer to 100% display of all these than many other technologies can provide). Screening this against alpha-1-acid-glycoprotein, a well-known carrier in the bloodstream, gave one pair that was significantly enhanced. Neither of the components had activity on their own. They then linked the two pharmacophores together via an 8-mer modified nucleic acid strand, and this (fluorescent) construct showed 190 nM binding, while leaving off either pharmacophore killed off the activity completely. They then linked the two with a lysine/PEG spacer, and found that this species was about 6 micromolar by ITC, with the control single-pharmacophore species showing no binding at all.

The other example is against carbonic anhydrase IX, a potential target and marker in some tumor types. The screen gave a 200x enrichment for one particular combination (totally different than the AGP screen, of course). They linked them together with the modified nucleic acids, finding that one of the components had a lot of the binding, but that the combination of the two was better still. They then tried several non-DNA linkers, settling on a two-aspartyl system that gave sub-nanomolar affinity in an FP assay, and was suitable for cell and in vivo applications as a potential biomarker imaging agent.

So one thing you can say is that the "find a new linker" problem is still with us, because these two results indicate that the team spent a good amount of time on that part of the process. On the other hand, though, you're still in a better situation than before, because you've already had to prepare "linkable" versions of your pharmacophores to get them onto the DNA strands, and you know from the assay that having stuff conjugated to the small molecules at these positions is acceptable. That's a big step, and surely speeds up the process of finding a new linked species. I look forward to seeing what Neri and Scheuermann make of this, and what others might as well. There are plenty of good targets out there for such a technique, of that there is no doubt.

Behind-the-scenes blogging note: the length of this post was provided to the readership courtesy of an hour-long delay on the inbound MBTA train to Boston!

Comments (1) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

February 4, 2015

The Old Binding Mode Switcheroo

Email This Entry

Posted by Derek

When a medicinal chemist gets a promising lead compound, the first thing that's done to it is to start making modifications to the structure. Add methyls or fluorines, take some of the substituents back off, switch stereochemistry, switch carbons and nitrogens around, that sort of thing. And there you start to build up a structure-activity relationship.

Maybe. Sometimes, though, the SAR doesn't seem to make a whole lot of sense. And in these cases, it may be that the underlying assumption of SAR is the problem. That is, we assume that the starting scaffold binds in pretty much the same way no matter what we do to it. There's a single binding mode, and some things make it better, and some things make it worse. But that's not always true. Fragment-based drug discovery illustrates that quite often, with minor changes to fragment-sized compounds leading to new (and often rather hard to predict) orientations in the binding site.

This new paper in Angewandte Chemie is a great illustration of just how weird things can get. The authors, from Marburg, were investigating a series o endothiapepsin inhibitors, with variations as shown at left. But when they did that ester-amide switch, X-ray crystallography showed a new binding mode. In fact, out of 8 compounds, they had four different binding, with the starting material the only singleton in the bunch. More potent compounds could be found in more than one series; it's not that any one particular mode was noticeably better.

The current example highlights the complexity of binding events and their strong dependence on seemingly minor effects of scaffold decoration and modifications. That 1, 7, 8, and 9 do not adopt the same binding geometry as observed in the green cluster can be rationalized retrospectively, but without the crystal structure determinations the adopted binding modes would have been difficult to predict, and, if so, would have hardly been believed or attracted sustained attention without our experimental evidence

The odds of getting so many different modes in so few compounds are not high, but they sure aren't zero, either. That's the question: just how often does this sort of thing go on, but pass unnoticed, written off as "just one of those things"? You can't always get X-ray structures of everything, but the authors here note that some discrepancies between affinity in the assay and those found by a thermal-shift assay seemed (in retrospect) to indicate some of the binding mode changes. So that might be a way to spot them. Not everyone will believe you if you keep using this an excuse, but it could be more valid than people realize. . .

Comments (10) + TrackBacks (0) | Category: Drug Assays | In Silico

February 2, 2015

More Thiol-Reactive PAIN Compounds

Email This Entry

Posted by Derek

Here's an open-access paper at J. Med. Chem. on still more classes of compounds that you're going to want to look out for. The authors, from the Mayo Clinic, the University of Minnesota, and AstraZeneca, ran a screen for an potential antifungal target, the Rtt109 histone acyltransferase enzyme. This screen used the fluorescent maleimide CPM as a chemical probe, and ran into trouble:

PAINS were computationally filtered at the beginning of our triage and were not initially evaluated in our post-HTS counter-screens. Post-HTS triage of approximately 1.5K primary screening hits demonstrated only a few confirmed actives. In retrospect, this indicated a significant portion of the screening hits were either false-positives or assay artifacts resulting from fluorescence quenching, compound-reagent interference and/or other mechanisms. Based on the chemical structures of the triaged compounds, we speculated many chemotypes from the primary HTS campaign were reacting with the CoA by-product to produce an interfering assay readout mimicking enzymatic inhibition. We also speculated many of these thiol-reactive compounds, including several series of PAINS we had previously triaged, could also inhibit enzymatic activity by reacting with protein cysteines, a recognized source of promiscuous enzyme inhibition and metabolic liability. We observed chemotypes that were enriched among the actives that appeared chemically similar to many of the published PAINS substructures, but were not flagged by our cheminformatic PAINS filters. . .The observation that some PAINS-like compounds may escape cheminformatics filters may have significant consequences for screening centers without experienced HTS triage personnel, especially if those performing HTS triage are overly reliant on (such) filters

That's the heart of the problem, for sure. High-throughput enzyme assays are is like any other assay technique: if you haven't done a lot of them, you run the risk of overvaluing what they tell you. In this case, they found a whole list of nasty mechanisms, including "addition-elimination reactions, nucleophilic aromatic substitution, buffer instability, disulfide bond formation and H2O2 production." That's enough to set off a lot of wild goose chases. In this case, out of 1500 putative hits, only three compounds were real. Imagine being faced with that situation if you're not alert for what could go wrong!
The compound classes they found, some of which are shown here, are not very promising looking even on first principles, at least to my eyes. But my eyes aren't everyone's, and neither am I always right about these things (as a glance at the structures of some big-selling drugs will remind me). So I'd recommend this paper to anyone embarking on HTS campaigns, especially anyone who's just getting into the area or is unsure about how to deal with a new library of unusual structures. The counterscreens and detection techniques detailed here are very useful things to keep in mind, and would do the drug-assay literature a lot of good if they were more widely heeded. Overall, the recommendation is to take an experimentalist, chemocentric view of hit triage, and I couldn't agree more.

Comments (8) + TrackBacks (0) | Category: Drug Assays

January 30, 2015

Underpowered And Overinterpreted

Email This Entry

Posted by Derek

Time for another "watch those statistics" post. I did one about this time last year, and I could do one every couple of months, to be honest. Here's a good open-access paper from the Royal Society on the problem of p-values, and why there are so many lousy studies out there in the literature. The point is summed up here:

If you use p=0.05 to suggest that you have made a discovery, you will be wrong at least 30% of the time. If, as is often the case, experiments are underpowered, you will be wrong most of the time.

True, true, and true. If you want to keep the false discovery rate down to below 5%, the paper says, you should be going for p<0.001. And just how many studies, of all kinds, across all fields, hit that standard? Not too damn many, which means that the level of false discovery out there is way north of 5%.

(This paper) deals only with the very simplest ideal case. We ask how to interpret a single p-value, the outcome of a test of significance. All of the assumptions of the test are true. The distributions of errors are precisely Gaussian and randomization of treatment allocations was done perfectly. The experiment has a single pre-defined outcome. The fact that, even in this ideal case, the false discovery rate can be alarmingly high means that there is a real problem for experimenters. Any real experiment can only be less perfect than the simulations discussed here, and the possibility of making a fool of yourself by claiming falsely to have made a discovery can only be even greater than we find in this paper.

The author of this piece is David Coqulhoun, a fact that some people will have guessed already, because he's been beating on this topic for many years now. (I've linked to some of his prickly opinion pieces before). He's not saying something that a lot of people want to hear, but I think it's something that more people should realize. A 95% chance of being right, across the board, would be a high standard to aim for, possibly too high for research to continue at a useful pace. But current standards are almost certainly too low, and we especially need to look out for this problem in studies of large medical significance.

Update: what this post needed was this graphic from XKCD!

Comments (50) + TrackBacks (0) | Category: Clinical Trials | Drug Assays | General Scientific News | The Scientific Literature

January 20, 2015

A New Cell Assay -And What It Says About Stapled Peptides

Email This Entry

Posted by Derek

I'm always looking out for new assays that might tell us what the heck is going on inside cells, so this paper caught my eye. The authors describe a new luciferase-based complementation assay for detecting protein-protein interactions. There are several things like this in the literature already (and for sale, too), but this one has what looks like a robust way to get the split-luciferase proteins expressed, and it seems to be picking up weaker and more transient interactions than most. For example, it's shown to pick up a specific ubiquitin ligase PPI that had been demonstrated by yeast two-hybrid assays, but never in living cells. (Depending on the signal/noise, this sensitivity could either be a bug or a feature!)

They also used this system on interactions of p53 (which has a good number of them), and found something interesting. The only-a-mother-could-love-it small molecule Nutlin 3a is believed to be an inhibitor of the p53-MDM2 interaction, but (as the current paper points out), this hasn't been conclusively demonstrated in living cells. This assay, though, confirmed that ". . .small-molecule PPI antagonists such as Nutlin-3a can selectively and rapidly disrupt preformed p53-Mdm2 complexes in living cells." But another reported PPI compound, SJ-172550, failed to show activity (its mechanism had already been reported as not just straight inhibition of the protein-protein interaction). RO-5963, another compound in this space, fared a bit better, but had a noticeably different profile than Nutlin-3a, which does argue for the ability of this assay to pick up fine details.

Stapled peptides have been used to target some of these p53 interactions as well, but conflicting data exist about just how well those work in this case. And the conflict continues: this assay showed some activity for ATSP-7041, but two other stapled peptides from the literature, SAHp53-8 and sMTide-02 (from that same ACS Chem Bio paper linked above) "exhibited no detectable ability to disrupt p53-Mdm2 or p53-Mdm4 complexes in living cells."

What they found, on further study, was that these stapled peptides seem to be cytotoxic, via some mechanism that has nothing to do with p53, and that this activity is inhibited by the presence of serum in the assay conditions. A cell-free assay system was developed, which indicated that the two problematic peptides were indeed able to disrupt the p53/MDM interactions as advertised - when they can get to it, that is. Adding serum to these assays did not disrupt things, though, which takes care of the possibility that something in serum just binds the peptides and keeps them from doing their thing.

So that leaves the serum as doing something else to keep the stapled species from actually entering the cells. What seems to be happening, from further experiments, is that the compounds are actually damaging cell membranes, which gives them a chance to get in and show activity under serum-free conditions. Adding 10% serum to the assay, though, seemed to protect the membranes from disruption (and thus makes the compounds show as inactive in the resulting cell assay). This effect was seem on plain old fibroblast cells as well, so it's not specific to cancer cells. And it wasn't seen with all stapled peptides, either - mutant forms of these very ones, for example, didn't have the same effect. Nutlin-3a didn't have it, either.

The authors suggest that this might be the source of some of the conflicting data in the literature on the effects of stapled peptide compounds, especially in this p53 area. People had noted some serum effects and cytotoxicity before, but much of this was explained via p53-dependent mechanisms. What this work shows is that the membrane damage is intrinsic to some of these peptides, and that this is going to have to be taken into account in future cell assays across the field. There are, of course, plenty of nonstapled peptides that are capable of causing membrane damage (some of them on purpose, as in natural antibiotic peptides), so this doesn't mean that stapled peptides are universally trouble. What it does mean, though, is that measurements of their penetration into cells have just gotten more complicated.

Comments (5) + TrackBacks (0) | Category: Biological News | Chemical Biology | Drug Assays | Toxicology

January 15, 2015

Aminothiazoles: A Yellow Caution Light

Email This Entry

Posted by Derek

I can agree with this one: "Promiscuous 2-aminothiazoles: A Frequent Hitting Scaffold". The authors (the med-chem group from Monash University in Australia, with a collaborator from AstraZeneca) make the case that the title compounds are usually more trouble than they're worth. They noticed that the 4-phenyl derivative went fourteen for fourteen in a set of diverse fragment screens, which certainly make a person wonder.

I've seen some similar behavior in that series, and I think that a lot of people have the impression that aminothiazoles hit pretty often in HTS campaigns. Indeed, there are several varieties of them in the canonical PAINs list. This paper investigates a range of substitution patterns, and finds that, far too often, they tend to hit a wide variety of protein targets. There's no clear SAR to get you out of that problem, either, which mirrors the general experience with PAIN-type compounds of flat SAR with no obvious trends. The binding is not due to impurities, nor is it covalent - it seems to be legitimate interaction with the protein, but there's just too damn much of it. In fact, several of these compounds seem to be binding at multiple sites per protein, which will certainly makes things hard to interpret.

But at the same time, you can find aminothiazoles in several marketed drugs, so you can't just rule the entire class out by saying that it never leads anywhere. As with many of the PAIN scaffolds, the best way to characterize the problem might be to say that although such motifs may appear in marketed drugs, your chances of success are likely to diminish if you try to follow their lead. The binding modes can make assay and SAR data hard to interpret during the early stage of optimization, and the over-friendly nature of the binding groups will always be a worry as you go on to cell assays or into animal models. You can press on, but your life has become more complicated. If you have something else equally interesting to work on that doesn't have these issues around it, you might be better off heading in that direction.

And to reiterate something that's come up a lot around here recently, and no doubt will again, the key thing about any new papers that report such structures as hits is to see whether the authors recognize any of this. Too often, such compounds get reported as if they were just happy HTS hits with nothing to hide - perfectly legitimate tool compounds, or even candidates for preclinical optimization, why not? But if the people doing such work have missed out on the known issues around their chemical matter, what else have they missed?

So here's a friendly competition: let's see who can spot the first 2-aminothiazole paper of 2015, showing this as hits for some assay, and let's see if it mentions anything about their promiscuity. The real acid test will be in about a year, when everyone who's even writing a manuscript has had time to see this one, but people have had time to see the other PAINs paper, and what good has that done them?

Comments (11) + TrackBacks (0) | Category: Chemical News | Drug Assays

A Kinase Inhibitor Overview

Email This Entry

Posted by Derek

Here's a very worthwhile review of kinase inhibitors by Oliver Hantschel, specifically focusing on the things that you might not otherwise notice. There are plenty of kinase inhibitors available in the catalogs as research tools, and they're listed as "Inhibitor of XYZ Kinase", as if that's all that they do. But what's really going on?

Quite a bit, in some cases. Making a truly selective kinase inhibitor has never been easy, especially in some parts of the kinase landscape. Hantschel makes a point that is worth thinking about - that it was perhaps fortunate that Gleevec (imatinib) was the first approved drug in this class. This is one of the more selective compounds, and its side-effect profile is quite manageable. If some of the later compounds had instead come first, the field might have developed differently. (Keep in mind that for some time kinases were considered basically undruggable, since the binding sites were considered too similar and the likelihood of trouble too high. The only kinase inhibitors known tended to have funky structures and poor selectivity, which was a self-reinforcing situation. I think, though, that people were still surprised that some of the broader-spectrum inhibitors were tolerated, well after the field was an accepted area for drug discovery).

The paper goes over the modern techniques for profiling kinase selectivity, which now can run to assays using hundreds of kinases. At this level of detail, most of the marketed kinase inhibitors still inhibit a dozen different enzymes or so, which is worth thinking about when such compounds are used as research tools - plenty of wrong assignments have been proposed over the years, and plenty more are surely waiting to be made. Modeling can tell you about which kinases have the most similar binding sites to the main target, which is a good start. But there are several compounds that are known to strongly inhibit different kinases by binding in completely different fashions, and throwing open a computational search to pick up changes of that size is problematic. Better to do the actual screening, when possible, while remembering that in vitro ranking may not completely translate to what's actually going on the cells.

As experiences in the clinic have demonstrated, there are an awful lot of kinase mechanisms in vivo that we don't yet understand. Dosing these inhibitors in animals and humans has helped to uncover a lot of new biology, but people should keep in mind that this is what they're probably going to be doing with each new compound. Hantschel goes into a couple of good examples of this (RAF and JAK2) where the initial rationale for the inhibitors as drugs turned out to be only part of some much larger and more complicated stories.

So there's plenty to think about just inside the kinase universe. The review goes on, though, to detail the many and various off-target effects that have been seen. A variety of non-kinase proteins can be picked up, including bromodomains, nucleotide-processing enzymes, oxidoreductases, and more. Some of these could be relevant to the cellular phenotypes that could otherwise be ascribed to kinase inhibition. It's important to keep in mind that kinase inhibitors are, in many cases, targeting binding motifs that have been re-used a number of times in living systems. The worries from thirty years ago about the whole field being untouchable were overblown, but it doesn't mean that there aren't plenty of surprises waiting.

From a drug development standpoint, those surprises are a bug. From a basic research standpoint, though, they're a feature. There's really no way that we could have learned so much about all these pathways other than by developing all these compounds and trying to untangle the results they've produced, and the untangling is going to continue for quite some time to come.

Comments (6) + TrackBacks (0) | Category: Cancer | Drug Assays | Drug Industry History

January 12, 2015

Rhodanines Are Even Trickier Than You Thought

Email This Entry

Posted by Derek

Bad things get said around here about rhodanines. That chemical class tends to hit in a lot of assays, especially assays against hard targets, leading to thoughts like the ones I was expressing the other day and in this post as well.. The problems with these compounds are many (and are mentioned in those blog posts linked above). They include the meta-problem of the sheer number of targets that these compounds can hit (making them problematic as leads or tools). That promiscuity is due to a number of mechanisms - direct binding, photoreactivity, and now we can add hydrolysis to the list.

Here's a new paper from Nature that shows that some rhodanine-based activity against penicillin-binding proteins and beta-lactamases comes, not from the parent compound, but from a thioenolate hydrolysis product.

This was the unexpected result of X-ray crystallographic work on the lead series, and it makes things rather hard to interpret:

". . .there are concerns regarding the promiscuity of rhodanines. Some consider rhodanines to be promiscuous inhibitors dueto their frequent appearance in high-throughput screening.However, rhodanines have also been described as ‘privileged scaffolds’ for inhibitor development. In cellular/biological contexts, promiscuity can make it difficult to correlate biochemical and biological outcomes with rhodanines. An important outcome of our work is the addition of another layer of complexity to the interpretation of rhodanine-based inhibition. . ."

So if you see a screening paper cheerfully reporting a rhondanine-based hit, and they haven't checked this issue, or the other known ones, then you know that you are reading inadequate work. And you can wonder what else has been missed.

Comments (12) + TrackBacks (0) | Category: Chemical News | Drug Assays

January 7, 2015

More Odd Compounds

Email This Entry

Posted by Derek

When you look into the literature on small-molecule agents for really tricky targets, something stands out to medicinal chemists immediately: the structures start to get strange. Examples of this sort of thing are beyond counting, but this recent paper will serve as well as any. It's from a large multicenter academic team, and proposes several compounds as ligands for Bax, a protein of the Bcl family that's involved in apoptosis and is a potential target for a range of lung cancers. There's a particular serine residue whose phosphorylation has been shown to alter Bax function significantly, and the present paper notes that there may well be a small-molecule-sized pocket nearby.

The work starts off by computationally docking a large collection of molecules from the NCI database to a model of this pocket. 36 out of 300,000 were found to score well, so the team exposed Bax-expressing cells to all 36 of them and looked for effects. Three of the compounds showed apoptotic effects, which were less marked in lung cancer cell lines that expressed less Bax. The compounds were able to compete with a fluorescent protein that also binds to Bax, and this assay gave affinities of around 50 nanomolar. Update: see the comments. These compounds give weirdly similar data.

These results sound promising. But let's look at the structures of the compounds. SMBA1 is not very attractive, for sure. You'd want to make sure that any fluorescence assay involving it doesn't suffer from some sort of interference, because it sure looks like a compound that would make its presence known in the UV/Vis spectrum. I have no experience with these sorts of fluorenylidene phenols, but some stability and reactivity checks would not be out of place.
The problem is, SMBA2 and SMBA3 make that first compound look like ibuprofen. I know that the NCI compound collection has a lot of wooly stuff in it, but come on. These are formaldehyde/amine condensation products. In aqueous solution, they exist in various equilibria depending on pH, and such structures are known to be reactive reagents in organic chemistry. (See the Delépine reaction, among others). SMBA3, in fact, is an intermediate structure from that very reaction, as if someone were trying to synthesis allylamine. I would be very worried about exposing such compounds to cells - there are no guarantees that they're going to remain the way that they're drawn, and the species that form can be quite reactive. Figuring out what's really going on with them would be quite a job.

I can see no indication from the paper that any of this bothered anyone. The only characterization that seems to have been done with these compounds was some DLS, dynamic light scattering, to look for aggregation. Nothing wrong with checking that, but there's no other chemical characterization or purity assessment of the compounds at all, or at least none that I can see. I mean, they're probably what they say on the label, but who can tell? And even if they're pure, how do they behave under the assay conditions?

The problems don't end there. SMBA2 is said to have 57 nM affinity to Bax. But that's an extraordinary value for a compound with a molecular weight of 168 - that kind of ligand efficiency should make a person suspect covalency. Covalent compounds are not necessarily bad, but you do want to examine them closely to see if they're working on the target you think they're hitting. I don't think the paper has any tests for that, such as checking for time-dependent inhibition or looking at the isolated Bax protein from the assay by mass spec. Meanwhile, SMBA3 has 54 nM affinity, and is shown as having a molecular weight of 308. But that includes the iodide counterion, and that doesn't count: this thing is not floating around in cells, nor binding to its putative target, with its iodide partner. Its real molecular weight is about 181, and that also represents a wildly high ligand efficiency. (There's also the question of how a charged quaternary compound like this gets into the cells so easily, but that's another issue).

The docked structures of these three compounds with Bax are shown in the paper. The two formaldehyde/amine compounds are shown interacting with two Asp residues, and there had certainly better be some strong interactions if you're going to pull out potencies like these. But the positively charged nitrogen in SMBA3 would surely seek out a negatively charged Asp if this were the case, a classic salt bridge, but that's not what the docking shows. Meanwhile, SMBA1, the fluorenyl compound, is shown just sort of floating there in space, not doing anything with the Asp residues, and not showing any strong interactions that I can see from the figure at all. Even the phenol doesn't seem to be doing anything, unless that's a pi-stack with Phe176. These dockings structures do not, unfortunately, inspire confidence.

So in light of all these objections and complications, let me get back to the point from the first paragraph of this post. When you get weird-looking structures out of a screen for a difficult target, you can explain them two ways. One possibility is that such targets, and such binding sites, are not evolutionarily optimized to bind any particular small molecules, and that regular "drug-like" chemical matter should not then be expected to hit them any more than anything else. Odd sites, in this view, will generate odd molecules. The other possibility, though, is that these things are false positives. The ways in which compounds can fool you are legion - in this case, I've mentioned some of the possibilities already, but there are more where those came from.

One thing that's for sure about targets like this one is that their intrinsic screening hit rates are very low. So that means, necessarily, that if the false positive rate is at any realistic level, then the bulk of what you get out of a screen will be just that: false positives. The challenge in screening these things is to dig through the garbage heap in search of the few hits that might be real. It's not a lot of fun, in some cases, because the list can be long and playing endless whack-a-mole with the compounds on it can be wearying. But you have to give all your hits the same tough love, or you run a significant risk of wasting your time.

So yes, I think that the odds are good that the compounds reported in this paper are false positives. They clearly seem to have cellular effects, and these may well be mediated by the Bax pathway. But it's a big leap, at least for me, to believe that everything lines up the way it's supposed to here. Odds are that anything reported for this Bax binding site is a false positive, and (although I hate to say it) this paper hasn't convinced me that its authors have given this problem sufficient attention.

Comments (28) + TrackBacks (0) | Category: Academia (vs. Industry) | Cancer | Chemical Biology | Drug Assays

December 10, 2014

A New Effect From JAK Inhibition

Email This Entry

Posted by Derek

Here's an interesting paper for people into metabolic disease: the authors report that JAK1/3 inhibitors (such as ​tofacitinib) seem to cause human adipocytes to convert from their normal phenotype into brown fat, a different beast altogether. Brown adipocytes are thermoregulators, and burn off lipids at an impressive rate. Many are the therapeutic approaches in metabolism that have targeted this pathway, but no one's every gotten into the real world.

Reading this paper, I wonder if it has something to do with irisin, a hormone that's also been reported to mediate this white-to-brown conversion. (No mention of it in the paper, though). JAK inhibition, in this case, seems to activate the Hedgehog pathway, which mediates the switchover. Of course, it mediates much more besides that, which is the problem here. As the authors themselves say:

The utility of JAK inhibition as a therapeutic strategy for obesity is complicated by the well-described role of this signalling pathway in the immune system. In fact, ​tofacitinib is approved in the United States to treat rheumatoid arthritis. Thus, if one were to imagine targeting adipose tissue by in vivo administration of an IFN–JAK–STAT inhibitor or similar compound it would almost certainly need to be delivered locally and prevented from spreading systemically or alternatively targeted selectively to white adipocytes. One could also conceive of a cell-based therapy wherein JAK inhibition of patient-derived adipocytes ex vivo is followed by transplantation to treat obesity, but this therapeutic modality would need to overcome numerous and significant obstacles before becoming a reality. A further limitation of the current study is the lack of evidence that JAK inhibition would promote metabolic browning in vivo, in particular in humans where evidence supporting this phenomenon is scant. Thus, additional research is required before inhibition of IFN–JAK–STAT signalling could be used therapeutically for the treatment of metabolic disorders.

There is another take-away from this work that could be of broad use, though - it relies on human adipocytes derived from a stem cell line, which seem to be, as the authors put it, "inexhaustible and rapidly expandable source of human adipocytes for screening and downstream assays". I suspect we'll see plenty of further phenotypic screening with them. Oh, and just as an aside, given my earlier post this morning: this is just the sort of thing that one might expect a discovery platform like Recursion's might pick up, if they're lucky. And it's just the sort of thing that might, in turn, keep them from discovering ten drugs a year, too.

Comments (5) + TrackBacks (0) | Category: Diabetes and Obesity | Drug Assays

Ambition From Recursion

Email This Entry

Posted by Derek

Here's one to put in the "hubris" file - we'll have to wait to see whether or not it's followed by the traditional divine retribution. According to Drug Discovery & Development (note: corrected source), startup Recursion Pharmaceuticals, out of the University of Utah, states that they're going to develop 100 drugs in ten years.

That paragraph break was deliberate, to allow the biopharma industry readers a chance to catch their breath, and perhaps find some paper towels for their keyboards. How, exactly, are these folks going to do this, you ask? The idea is not crazy, nor is it stupid - they're going to repurpose known drugs through phenotypic screening. Not a bad plan at all. In fact, it's so reasonable that a lot of other people have had it. (It's not clear from the article whether or not Recursion's people are completely aware of this - there's quite a bit of stuff contrasting their approach to traditional target-based drug discovery). Where they feel they have an edge is in their computational approach to high-content screening:

". . .most drug-repurposing successes happen serendipitously, or with an educated guess based on deep biological understanding of a given disease. And while that may work for well-studied diseases, we don’t have that same level of understanding with rare disorders,” said Li, Recursion’s co-founder. “With disease modeling and computational algorithms we developed, we’re able to make drug repurposing scalable for use with rare diseases.”

They actually do have a couple of leads for cerebral cavernous malformation, but the article mentions in passing that the Utah lab has been studying that particular condition for about ten years now. They seem to feel, though, that this experience should carry over to thousands of other rare-disease targets pretty easily, though. Recursion's co-founder and CEO Christopher Gibson, who has just finished his PhD, has anticipated my objections:

The idea of a pharmaceutical company tackling hundreds of diseases in a year is unthinkable for even the largest drug makers, which may be working on a dozen diseases at a time. “There will be doubters, those who say we’re naïve to think we can accomplish that. But that’s where science is taking is us,” Gibson said.

I actually appreciate the amount of nerve being shown here. True, I'd appreciate it more in someone who's actually had some more experience, because then it would really stand out. Drug discovery is a pretty humbling experience - at least in my experience, and I don't think it's just me. I think that "naïve" is a completely accurate description of Recursion's plan, to be honest. I believe that they truly don't know what they're in for, and that there may well be many things that they don't even know yet that they should know. That's naïveté, and no mistake.

But on the other hand, everyone starts out like this, more or less. I can't fault people who've never done this sort of thing just for having no experience. I suppose the difference is that we don't all start companies and tell everyone that we're going to discover one new drug every four weeks for the next ten years in a row, or make sure that we're quoted saying things like “These drugs are just sitting in freezers. We’re saying, ‘Give us your drugs and we’ll monetize them". Still, I think that Recursion should come on down and give it a try. Anyone with ideas about how to improve drug discovery should come on down and give it a try. Reality will sort us all out soon enough. Won't it just?

Comments (25) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development

December 4, 2014

Whatever Happened to the Fish and the Frogs?

Email This Entry

Posted by Derek

Eight or ten years ago, there was a good deal of excitement about non-mammalian small animal model systems for compound screening - specifically fish and frogs. More specifically, zebrafish and Xenopus. A number of small companies started up to do this sort of thing, and large companies paid attention as well. A correspondent, though, wrote me the other day noting that few (if any) of these companies seem to have made it. Phylonix and Znomics seem to be inactive, and InDanio, while apparently still with us, has a low profile. (Are there others?)

More generally, it's worth asking what's become of the whole idea. I've read some interesting papers over the years using these systems in compound screening, mostly uncovering effects developmental pathways. But how has it worked out overall? There are still people working in the field, of course, but have they been able to get any traction in drug discovery? From my perspective, I just know that I seem to hear a lot less about this than I did a few years back.

Is there still a zebrafish case to be made, or one for the (evolutionarily closer) Xenopus? If so, is that case best made for developmental biology/embryology, or for more open-ended high-content phenotypic screening, or for toxicology? Thoughts welcome.

Comments (23) + TrackBacks (0) | Category: Animal Testing | Drug Assays | Toxicology

October 20, 2014

Compound Properties: Starting a Renunciation

Email This Entry

Posted by Derek

I've been thinking a lot recently about compound properties, and what we use them for. My own opinions on this subject have been changing over the years, and I'm interested to see if I have any company on this.

First off, why do we measure things like cLogP, polar surface area, aromatic ring count, and all the others? A quick (and not totally inaccurate) answer is "because we can", but what are we trying to accomplish? Well, we're trying to read the future a bit and decrease the horrendous failure rates for drug candidates, of course. And the two aspects that compound properties are supposed to help with are PK and tox.

Of the two, pharmacokinetics is the one with the better shot at relevance. But how fine-grained can we be with our measurements? I don't think it's controversial to say that compounds with really high cLogP values are going to have, on average, more difficult PK, for various reasons. Compounds with lots of aromatic rings in them are, on average, going to have more difficult PK, too. But how much is "lots" or "really high"? That's the problem, because I don't think that you can draw a useful line and say that things on one side of it are mostly fine, and things on the other are mostly not. There's too much overlap, and too many exceptions. The best you can hope for, if you're into line-drawing, is to draw one up pretty far into the possible range and say that things below it may or may not be OK, but things above it have a greater chance of being bad. (This, to my mind, is all that we mean by all the "Rule of 5" stuff). But what good does that do? Everyone doing drug discovery already knows that much, or should. Where we get into trouble is when we treat these lines as if they were made of electrified barbed wire.

That's because of a larger problem with metrics aimed at PK: PK is relatively easy data to get. When in doubt, you should just dose the compound and find out. This makes predicting PK problems a lower-value proposition - the real killer application would be predicting toxicology problems. I fear that over the years many rule-of-five zealots have confused these two fields, out of a natural hope that something can be done about the latter (or perhaps out of thinking that the two are more related than they really are). That's unfortunate, because to my mind, this is where compound property metrics get even less useful. That recent AstraZeneca paper has had me thinking, the one where they state that they can't reproduce the trends reported by Pfizer's group on the influences of compound properties. If you really can take two reasonably-sized sets of drug discovery data and come to opposite conclusions about this issue, what hope does this approach have?

Toxicology is just too complicated, I think, for us to expect that any simple property metrics can tell us enough to be useful. That's really annoying, because we could all really use something like that. But increasingly, I think we're still on our own, where we've always been, and that we're just trying to make ourselves feel better when we think otherwise. That problem is particularly acute as you go up the management ladder. Avoiding painful tox-driven failures is such a desirable goal that people are tempted to reach for just about anything reasonable-sounding that holds out hope for it. And this one (compound property space policing) has many other tempting advantages - it's cheap to implement, easy to measure, and produces piles of numbers that make for data-rich presentations. Even the managers who don't really know much chemistry can grasp the ideas behind it. How can it not be a good thing?

Especially when the alternative is so, so. . .empirical. So case-by-case. So disappointingly back-to-where-we-started. I mean, getting up in front of the higher-ups and telling them that no, we're not doing ourselves much good by whacking people about aromatic ring counts and nitrogen atom counts and PSA counts, etc., that we're just going to have to take the compounds forward and wait and see like we always have. . .that doesn't sound like much fun, does it? This isn't what anyone is wanting to hear. You're going to do a lot better if you can tell people that you've Identified The Problem, and How to Address It, and that this strategy is being implemented right now, and here are the numbers to prove it. Saying, in effect, that we can't do anything about it runs the risk of finding yourself replaced by someone who will say that we can.

But all that said, I really am losing faith in property-space metrics as a way to address toxicology. The only thing I'm holding on to are some of the structure-based criteria. I really do, for example, think that quinones are bad news. I think if you advance a hundred quinones into the clinic, that a far higher percentage of them will fail due to tox and side effects than a hundred broadly similar non-quinones. Same goes for rhodanines, and a few other classes, those "aces in the PAINS deck" I referred to the other day. I'm still less doctrinaire about functional groups than I used to be, but I still have a few that I balk at.

And yes, I know that there are drugs with all these groups in them. But if you look at the quinones, for example, you find mostly cytotoxics and anti-infectives which are cytotoxins with some selectivity for non-mammalian cells. If you're aiming at a particularly nasty target (resistant malaria, pancreatic cancer), go ahead and pull out all the stops. But I don't think anyone should cheerfully plow ahead with such structures unless there are such mitigating circumstances, or at least not without realizing the risks that they're taking on.

But this doesn't do us much good, either - most medicinal chemists don't want to advance such compounds anyway. In fact, rather than being too permissive about things like quinones, most of us are probably too conservative about the sorts of structures we're willing to deal with. There are a lot of funny-looking drugs out there, as it never hurts to remind oneself. Peeling off the outer fringe of these (and quinones are indeed the outer fringe) isn't going to increase anyone's success rate much. So what to do?

I don't have a good answer for that one. I wish I did. It's a rare case when we can say, just by looking at its structure, that a particular compound just won't work. I've been hoping that the percentages would allow us to say more than that about more compounds. But I'm really not sure that they do, at least not to the extent that we need them to, and I worry that we're kidding ourselves when we pretend otherwise.

Comments (32) + TrackBacks (0) | Category: Drug Assays | Drug Development | In Silico

October 17, 2014

Different Screening, Different Thermodynamics?

Email This Entry

Posted by Derek

Chris Lipinski and the folks at Collaborative Drug Discovery send word of an interesting webinar that will take place this coming Wednesday (October 22nd) at 2 PM EST. It's on enthalpic and entropic trends in ligand binding, and how various screening and discovery techniques might bias these significantly.

Here's the registration page if you're interested. I'm curious about what they've turned up - my understanding is that it will explore, among other things, the differences in molecules selected by industry-trained medicinal chemists versus the sorts that are reported by more academic chemical biologists. As has come up here several times in the past, there certainly do seem to be some splits there, and the CDD people seem to have some numbers to back up those impressions.

Comments (7) + TrackBacks (0) | Category: Academia (vs. Industry) | Chemical Biology | Drug Assays

October 15, 2014

Not The Sort Of Thing You'd Work With, Given a Choice

Email This Entry

Posted by Derek

Here's a paper that illustrates a different way of looking at the world than many medicinal chemists would have. It discusses inhibitors of SETD8, an unusual epigenetic enzyme that is the only known methyltransferase to target lysine 20 of histone H4. Inhibitors of it would help to unravel just what functions that has, presumably several that no other pathway is quite handling. But finding decent methyltransferase inhibitors has not been easy.
When you search for them, actually, you find compounds like the ones in this paper. Most medicinal chemists will look at these, say the word "quinone", perhaps take a moment spit on the floor or add a rude adjective, and move on to see if there's anything better to look at. Quinones are that unpopular, and with good reason. They're redox-active, can pick up nucleophiles as Michael acceptors, react with amines - they have a whole suite of unattractive behaviors. And that explains their profile in cells and whole animals, with a range of toxic, carcinogenic, and immunologic liabilities. A lot of very active natural products have a quinone in them - it's a real warhead. No medicinal chemist with any experience would feel good about trying to advance one as a lead compound, and (for the same reasons) they tend to make poor tool compounds as well. You just don't know what else they're hitting, and the chance of them hitting something else are too high.

The authors of this paper, though, have a higher tolerance:

In the present work, we characterize these compounds and demonstrate that NSC663284, BVT948, and ryuvidine (3 out of the 4 HTS hits) inhibit SETD8 via different modes. NSC663284 (SPS8I1), ryuvidine (SPS8I2), and BVT948 (SPS8I3) efficiently and selectively suppress cellular H4K20me1 at doses lower than 5 μM within 24 h. . . The cells treated with SPS8I1−3 (Small-molecule Pool of SETD8 Inhibitor) recapitulate cell-cycle-arrest phenotypes similar to what were reported for knocking down SETD8 by RNAi. Given that the three compounds have distinct structures and inhibit SETD8 in different manners, they can be employed collectively as chemical genetic tools to interrogate SETD8-involved methylation.

I would be very careful about doing that, myself. I don't find those structures as distinct as all that (quinone, quinone, quinone), and I'm not surprised to find that they arrest the cell cycle. But do they do it via SETD8? To be fair, they do show selectivity over the other enzymes used in the screening panel (SETD7, SETD2, and GLP). They went on to profile them against several lysine methyltransferases and several arginine methyltransferases. The most selective of the bunch was 2.5x more active against SETD8 compared to the next most active target, which is honestly not a whole lot. (And I note that the authors spent some time a few paragraphs before talking about how their activity measurements are necessarily uncertain).

They do address the quinone problem, but in a somewhat otherworldly manner:

Given that SPS8I1−3 are structurally distinct except for their quinonic moiety (Figure 1a, highlighted in red), we reasoned that they may act on SETD8 differently (e.g., dependence on cofactor or substrate). . .

This, to many medicinal chemists, is a bit like saying that several species of poisonous snake are distinct except for their venom-filled fangs. The paper does seem to find differences in how the three inhibitors respond to varying substrate concentrations, but they also find (unsurprisingly) that all three work by covalent inhibition. Studies on mutant forms of the enzyme suggest strongly that two of the compounds are hitting a particular Cys residue (270), while the third "may target Cys residues in a more general manner". To their credit, they did try three quinone-containing compounds from commercial sources and found them inactive, but that just shows that not every quinone inhibits their enzyme.

This, too, is what you'd expect: if you did a full proteome analysis of what a given quinone compound hits, I'm sure that you'd find varying fingerprints for each one. But even though I have no objection to covalent inhibitors per se, I'm nervous about ones that have so many potential mechanisms. The size and shape of the three compounds shown will surely keep them from doing all the damage that a smaller quinone is capable of doing, but I fear that there's still plenty of damage in them.

Indeed, when they do cell assays, they find that each of the compounds has a somewhat different profile of cell cycle arrest, and say that this is probably due to their off-target effects. But they go on to wind things up like this:

Structurally distinct SPS8I1−3 also display different modes of SETD8 inhibition. Such differences also make SPS8I1−3 less likely to act on other common cellular targets besides SETD8. As a result, the shared phenotypes of the 3 compounds are expected to be associated with SETD8 inhibition. . .Such robust inhibition of SETD8 by SPS8I1−3, together with their different off-target effects, argues that these compounds can be used collectively as SETD8 inhibitors to offset off-target effects of individual reagents. At this stage, we envision using all three compounds to examine SETD8 inhibition and then focusing on the phenotypes shared by all of them.

I have to disagree there. I would be quite worried about how many other cellular processes are being disrupted by these compounds. In fact, the authors already point to some of these. Their SPS8I1, they note, has already been reported as a CDC25 inhibitor. SPS8I2 has been shown to be a CDK2/4 inhibitor, and SPS8I3 has been reported as an inhibitor of a whole list of protein tyrosine phosphatases. None of these enzymes, I would guess, has any particular great structural homology with SETD8, and those activities are surely only the beginning. How is all this to be untangled? Using all three of them to study the same system is likely to just confuse things more rather than throwing light on common mechanisms. Consider the background: even a wonderful, perfectly selective SETD8 inhibitor would be expected to induce a complex set of phenotypes, varying on the cell type and the conditions.

And these are not wonderful inhibitors. They are quinones, aces in the deck of PAINS. No matter what, they need a great deal more characterization before any conclusions can be drawn from their activity. A charitable view of them would be that such characterization, along with a good deal of chemistry effort, might result in derivatives that have a decent chance of hitting SETD8 in a useful manner. An uncharitable view would be that they should be poured into the red waste can before they use up any more time and money.

Comments (42) + TrackBacks (0) | Category: Drug Assays

October 14, 2014

Combichem Into Drugs: How Many?

Email This Entry

Posted by Derek

So here's a question I got from a reader the other day, that I thought I'd put up on the site. How many drugs have there been whose origins were in combichem? I realize that this could be tricky to answer, because compound origins are sometimes forgotten or mysterious. But did the combichem boom of the 1990s produce any individual compound success stories?

Comments (49) + TrackBacks (0) | Category: Drug Assays | Drug Industry History

October 6, 2014

A New Way to Estimate a Compound's Chances?

Email This Entry

Posted by Derek

Just a few days ago we were talking about whether anything could be predicted about a molecule's toxicity by looking over its biophysical properties. Some have said yes, this is possible (that less polar compounds tend to be more toxic), but a recent paper has said no, that no such correlation exists. This is part of the larger "Rule of 5" discussion, about whether clinical success in general can be (partially) predicted by such measurements (lack of unexpected toxicity is a big factor in that success). And that discussion shows no sign of resolving any time soon, either.

Now comes a new paper that lands right in the middle of this argument. Douglas Kell's group at Manchester has analyzed a large data set of known human metabolites (the Recon2 database, more here) and looked at how similar marketed drugs are to the structures in it. Using MACCS structural fingerprints, they find that 90% of marketed drugs have a Tanimoto similarity of more than 0.5 to at least one compound in the database, and suggest that this could be a useful forecasting tool for new structures.

Now, that's an interesting idea, and not an implausible one, either. But the next things to ask are "Is it valid?" and "What could be wrong with it?" That's the way we learn how to approach pretty much anything new that gets reported in science, of course, although people do tend to take it the wrong way around the dinner table. Applying that in this case, here's what I can think of that could be off:

1. Maybe the reason that everything looks like one of the metabolites in the database is that the database contains a bunch of drug metabolites to start with, perhaps even the exact ones from the drugs under discussion? This isn't the case, though: Recon2 contains endogenous metabolites only, and the Manchester group went through the list removing compounds that are listed as drugs but are also known metabolites (nutritional supplements, for the most part).

2. Maybe Tanimoto similarities aren't the best measurement to use, and overestimate things? Molecular similarity can be a slippery concept, and different people often mean different things by it. The Tanimoto coefficient is the ratio of shared features of two molecules to their unique features, so a Tanimoto of 1 means that the two are identical. What does a coefficient of 0.5 tell us? That depends on how those "features" are counted, as one could well imagine, and the various ones are usually referred to as compound "fingerprints". The Manchester group tried several of these, and settled on the 166 descriptors of the MACCS set. And that brings up the next potential problem. . .

3. Maybe MACCS descriptors aren't the best ones to use? I'm not enough of an informatics person to say, although this point did occur to the authors. They don't seem to know the answer, either, however:

However, the cumulative plots of the (nearest metabolite Tanimoto similarity) for each drug using different fingerprints do differ quite significantly depending on which fingerprint is used, and clearly the well-established MACCS fingerprints lead to a substantially greater degree of ‘metabolite-likeness’ than do almost all the other encodings (we do not pursue this here).

So this one is an open question - it's not for sure if there's something uniquely useful about the MACCS fingerprint set here, or if there's something about the MACCS fingerprint set that makes it just appear to be uniquely useful. The authors do note in the paper that they tried to establish that the patterns they saw were ". . .not a strange artefact of the MACCS encoding itself." And there's another possibility. . .

4. Maybe the universe of things that make this cutoff is too large to be informative? That's another way of asking "What does a Tanimoto coefficient of 0.5 or greater tell you?" The authors reference a paper (Baldi and Nasr) on that very topic, which says:

Examples of fundamental questions one would like to address include: What threshold should one use to assess significance in a typical search? For instance, is a Tanimoto score of 0.5 significant or not? And how many molecules with a similarity score above 0.5 should one expect to find? How do the answers to these questions depend on the size of the database being queried, or the type of queries used? Clear answers to these questions are important for developing better standards in chemoinformatics and unifying existing search methods for assessing the significance of a similarity score, and ultimately for better understanding the nature of chemical space.

The Manchester authors say that applying the methods of that paper to their values show that they're highly significant. I'll take their word for that, since I'm not in a position to run the numbers, but I do note that the earlier paper emphasizes that a particular Tanimoto score's significance is highly dependent on the size of the database, the variety of molecules in it, and the representations used. The current paper doesn't (as far as I can see) go into the details of applying the Baldi and Nasr calculations to their own data set, though.

The authors have done a number of other checks, to make sure that they're not being biased by molecular weights, etc. They looked for trends that could be ascribed to molecular properties like cLogP, but found none. And they tested their hypothesis by running 2000 random compounds from Maybridge through, which did indeed generate much different-looking numbers than the marketed drugs.

As for whether their overall method is useful, here's the Manchester paper's case:

. . .we have shown that when encoded using the public MDL MACCS keys, more than 90 % of individual marketed drugs obey a ‘rule of 0.5’ mnemonic, elaborated here, to the effect that a successful drug is likely to lie within a Tanimoto distance of 0.5 of a known human metabolite. While this does not mean, of course, that a molecule obeying the rule is likely to become a marketed drug for humans, it does mean that a molecule that fails to obey the rule is statistically most unlikely to do so.

That would indeed be a useful thing to know. I would guess that people inside various large drug organizations are going to run this method over their own internal database of compounds to see how it performs on their own failures and successes - and that is going to be the real test. How well it performs, though, we may not hear for a while. But I'll keep my ears open, and report on anything useful.

Comments (35) + TrackBacks (0) | Category: Drug Assays | In Silico

October 2, 2014

We Can't Calculate Our Way Out of This One

Email This Entry

Posted by Derek

Clinical trial failure rates are killing us in this industry. I don't think there's much disagreement on that - between the drugs that just didn't work (wrong target, wrong idea) and the ones that turn out to have unexpected safety problems, we incinerate a lot of money. An earlier, cheaper read on either of those would transform drug research, and people are willing to try all sorts of things to those ends.

One theory on drug safety is that there are particular molecular properties that are more likely to lead to trouble. There have been several correlations proposed between high logP (greasiness) and tox liabilities, multiple aromatic rings and tox, and so on. One rule proposed in 2008 by a group at Pfizer is that clogP >3 and total polar surface area less than 75 square angstroms is a good cutoff - compounds on the other side of it are about 2.5 times more likely to run into trouble. But here's a paper in MedChemComm that asks if any of this has any validity:

What is the likelihood of real success in avoiding attrition due to toxicity/safety from using such simple metrics? As mentioned in the beginning, toxicity can arise from a wide variety of reasons and through a plethora of complex mechanisms similar to some of the DMPK endpoints that we are still struggling to avoid. In addition to the issue of understanding and predicting actual toxicity, there are other hurdles to overcome when doing this type of historical analysis that are seldom discussed.

The first of these is making sure that you're looking at the right set of failed projects - that is, ones that really did fail because of unexpected compound-associated tox, and not some other reason (such as unexpected mechanism-based toxicity, which is another issue). Or perhaps a compound could have been good enough to make it on its own under other circumstances, but the competitive situation made it untenable (something else came up with a cleaner profile at about the same time). Then there's the problem of different safety cutoffs for different therapeutic areas - acceptable tox for a pancreatic cancer drug will not cut it for type II diabetes, for example.

The authors did a thorough study of 130 AstraZeneca development compounds, with enough data to work out all these complications. (This is the sort of thing that can only be done from inside a company's research effort - you're never going to have enough information, working from outside). What they found, right off, was that for this set of compounds the Pfizer rule was completely inverted. The compounds on the too-greasy side actually had shown fewer problems (!) The authors looked at the data sets from several different angles, and conclude that the most likely explanation is that the rule is just not universally valid, and depends on the dataset you start with.

The same thing happens when you look at the fraction of sp3 carbons, which is a characteristic (the "Escape From Flatland" paper) that's also been proposed to correlate with tox liabilities. The AZ set shows no such correlation at all. Their best hypothesis is that this is a likely correlation with pharmacokinetics that has gotten mixed in with a spurious correlation with toxicity (and indeed, the first paper on this trend was only talking about PK). And finally, they go back to an earlier properties-based model published by other workers at AstraZeneca, and find that it, too, doesn't seem to hold up on the larger, more curated data set. Their-take home message: ". . .it is unlikely that a model of simple physico-chemical descriptors would be predictive in a practical setting."

Even more worrisome is what happens when you take a look at the last few years of approved drugs and apply such filters to them (emphasis added):

To investigate the potential impact of following simple metric guidelines, a set of recently approved drugs was classified using the 3/75 rule (Table 3). The set included all small molecule drugs approved during 2009–2012 as listed on the ChEMBL website. No significant biases in the distribution of these compounds can be seen from the data presented in Table 3. This pattern was unaffected if we considered only oral drugs (45) or all of the drugs (63). The highest number of drugs ends up in the high ClogP/high TPSA class and the class with the lowest number of drugs is the low ClogP/low TPSA. One could draw the conclusion that using these simplistic approaches as rules will discard the development of many interesting and relevant drugs.

One could indeed. I hadn't seen this paper myself until the other day - a colleague down the hall brought it to my attention - and I think it deserves wider attention. A lot of drug discovery organizations, particularly the larger ones, use (or are tempted to use) such criteria to rank compounds and candidates, and many of us are personally carrying such things around in our heads. But if these rules aren't valid - and this work certainly makes it look as if they aren't - then we should stop pretending as if they are. That throws us back into a world where we have trouble distinguishing troublesome compounds from the good ones, but that, it seems, is the world we've been living in all along. We'd be better off if we just admitted it.

Comments (25) + TrackBacks (0) | Category: Drug Assays | Drug Development | In Silico | Toxicology

September 26, 2014

PAINS Go Mainstream

Email This Entry

Posted by Derek

Well, I'm back in the Eastern Time Zone after flying in from Basel (and Amsterdam) yesterday. And the first thing I wanted to mention was this article from Jonathan Baell and Michael Walters inNature, on the PAINS compounds. It's good to see the journal cover this issue (and I was impressed that they got New Yorker cartoonist Roz ChastRoz Chast to illustrate it).

PAINS are, of course, nasty frequent-hitting compounds that should be approached with great caution in any sort of screen for activity. This topic has come up many times on the blog (for someone writing about chemistry and drug discovery, there's no way it couldn't have), most recently just a few weeks ago. There are a lot of these things out in the literature (and the catalogs), and they just keep on coming. Now a wider audience gets to hear about the problem:

Academic researchers, drawn into drug discovery without appropriate guidance, are doing muddled science. When biologists identify a protein that contributes to disease, they hunt for chemical compounds that bind to the protein and affect its activity. A typical assay screens many thousands of chemicals. ‘Hits’ become tools for studying the disease, as well as starting points in the hunt for treatments.

But many hits are artefacts — their activity does not depend on a specific, drug-like interaction between molecule and protein. A true drug inhibits or activates a protein by fitting into a binding site on the protein. Artefacts have subversive reactivity that masquerades as drug-like binding and yields false signals across a variety of assays.

That's the problem, all right. It's not like ugly-looking compounds can never become drugs, and it's not like they can't be starting points for research. But the odds are against them, and you have to realize that, and you also have to realize why this "hit" you've just uncovered may well be spurious (at worst) or need a lot of extra work (at best). Far, far too many papers from less experienced research teams seem to be oblivious to these concerns. Compound hits? Compound good!

Appropriately, this piece calls out the rhodanines as perfect examples of the problem:

Rhodanines exemplify the extent of the problem. A literature search reveals 2,132 rhodanines reported as having biological activity in 410 papers, from some 290 organizations of which only 24 are commercial companies. The academic publications generally paint rhodanines as promising for therapeutic development. In a rare example of good practice, one of these publications (by the drug company Bristol-Myers Squibb) warns researchers that these types of compound undergo light-induced reactions that irreversibly modify proteins. It is hard to imagine how such a mechanism could be optimized to produce a drug or tool. Yet this paper is almost never cited by publications that assume that rhodanines are behaving in a drug-like manner.

Very occasionally, a PAINS compound does interact with a protein in a specific drug-like way. If it does, its structure could be optimized through medicinal chemistry. However, this path is fraught — it can be difficult to distinguish when activity is caused by a drug-like mechanism or something more insidious. Rhodanines also occur in some 280 patents, a sign that they have been selected for further drug development. However, to our knowledge, no rhodanine plucked out of a screening campaign is in the clinic or even moving towards clinical development. We regard the effort to obtain and protect these patents (not to mention the work behind them) as a waste of money.

Yeah, I wouldn't spend much on trying to stake a claim to these things, either. If you haven't done much screening, you may not appreciate just how many false positives are out there (and for difficult targets, how few real positives there may be). I see people in the literature screening little libraries of a few thousand compounds from a catalog and reporting hit after hit, even in very tricky systems, while in industry we're used to running hundreds of thousands of compounds past some of these things and coming up with squat. Well, after checking the "hits" for purity, aggregation behavior, reactivity, and profiles from past screening campaigns, that is.

Here's the sad truth: If you're doing a small-molecule screen to affect transcription factors, protein-protein targets, or anything in general that doesn't have an evolutionary optimized small-molecule binding site, you'd better assume that the vast majority of any hits you get are false positives. There's almost no way that they can be anything else. The true hit rate for some of these things against any sort of typical compound collection is damn near zero, which means that the ways your compounds can be wrong far outnumber the ways that they can be right.

Every single hit, for any assay, should be regarded with appropriate suspicion. Purity check first, LC/MS and NMR. Is it what it says on the label? You might be surprised how often it isn't (or isn't any more, even if it started out OK). If you have solid material and DMSO stock, check both of them, because things diverge on storage. It's a very good idea to take your interesting hits, run them through a plug of silica gel, and test them again. That's especially true if they have any color to them (but keep in mind, some assay-killing contaminants are completely colorless). The gold standard is resynthesis: if you can make the compound again and purify it, and it still works, you at least know you can trust it that far. If you can't, well, how exactly is this compound going to do anyone any good?

Note that we haven't even gotten to the PAINS yet. There are a lot of clean, accurately labeled compounds that should be chucked into the waste can, too, which is where the Baell PAINS list comes in. You're going to want to check for aggregation: run your assay with some detergent in it, or do some dynamic light scattering or any of several other techniques. A lot of false-positive compounds are aggregators, and you can't completely predict which ones they might be (it varies according to assay conditions).

You're also going to want to run your hits through some other assays. How promiscuous are they? If you have access to data from multiple screening campaigns with the same compound collection, good for you. If you don't, you should strongly consider sending your hot compound(s) out for a commercial screening panel. Don't just pick the similar targets to screen - you want those, of course, but you want all kinds of other stuff. If a compound hits against widely disparate protein classes, it's a PAIN, and is set to cause trouble. Don't assume that they're clean - don't assume that any compound is clean, because it almost certainly isn't. That goes for marketed drugs, too - the question is, does it have selectivity that you can live with, or not?

Those are the big tests, and believe me, they'll clear out your initial list of screening hits for you. If your target is a tough one to start with, they may well clear out everything. Better that, though, than working on (and publishing) crap.

Comments (15) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays

September 23, 2014

Thoughts on Fragment-Based Drug Discovery

Email This Entry

Posted by Derek

I've been enjoying the FBLD fragment conference in Basel. There have been many good talks, and it's been instructive to talk shop with people as well. Some things that various participants (and I) have noted:

(1) There are a lot of industry people here, from all over. Fragment-based methods have clearly made a big impression across drug discovery - academia finds it a low-barrier way to get into compound screening, and the industrial groups clearly find it useful as well. This has already lasted longer than the combichem boom of the 1990s (a point I'll be mentioning in my talk here tomorrow), and at this point, it would appear to be not a fad at all. Fragment-derived compounds are marching along through the clinic as we speak.

(2) The number of instrumental and biophysical techniques to do fragment work is still growing. There are always NMR screens, SPR, and X-ray crystallography (and many variations on each of these), calorimetry and thermal shift experiments, and so on. But there are mass spec methods, chromatographic ones, thermophoresis and electrophoresis, and more where those came from. Which is good - as anyone who'd done this can tell you, you need orthogonal ways of looking at the compounds.

(3) Several of these technique are, though, still a bit "operator-dependent". SPR, just to pick one, needs someone experienced at the controls for the trickier experiments, because there are a lot of things that can give you funny-looking data (aggregation, compounds that bind to the chip surface, super-stoichiometric binding, and more). It's not a good part-time occupation. You can say the same thing about a lot of other screening techniques that are used for traditional high-throughput screening, though, as anyone who's troubleshooted FRET, FP, or AlphaScreen assays will tell you at length.

(4) There are still a lot of important and interesting things we don't quite know about fragment collections and fragment binding. Some of these are just beginning to get sorted out - what are the physical characteristics of a good fragment screening set (beyond the obvious ones of size and solubility)? Are there different sorts of collections that could give you better hit rates against targets like protein-protein interactions (there have been many examples of these at the meeting). What are the relationships between selectivity at the fragment level and selectivity in the eventual optimized compounds? The answers the questions like these are still being written.

I have more thoughts, naturally, but some of those my employer has first call on. I have a talk to give tomorrow (one of those 30,000-foot-overview things), right at the end of the conference, and I'll probably unburden myself of a few opinions then.

Comments (9) + TrackBacks (0) | Category: Chemical News | Chemical News | Drug Assays

September 8, 2014

Reactive Groups: Still Not So Reactive

Email This Entry

Posted by Derek

Just how reactive are chemical functional groups in vivo? That question has been approached by several groups in chemical biology, notably the Cravatt group at Scripps. One particular paper from them that I've always come back to is this one, where they profiled several small electrophiles across living cells to see what they might pick up. (I blogged about more recent effort in this vein here as well).

Now there's a paper out in J. Med. Chem. that takes a similar approach. The authors, from the Klein group at Heidelberg, took six different electrophiles, attached them to a selection of nonreactive aliphatic and aromatic head groups, and profiled the resulting 72 compounds across a range of different proteins. There are some that are similar to what's been profiled in the Cravatt papers and others (alpha-chloroketones), but others I haven't seen run through this sort of experiment at all.

And what they found confirms the earlier work: these things, even fairly hot-looking ones, are not all that reactive against proteins. Acrylamides of all sorts were found to be quite clean, with no inhibition across the enzymes tested, and no particular reaction with GSH in a separate assay. Dimethylsulfonium salts didn't do much, either (although a couple of them were unstable to the assay conditions). Chloroacetamides showed the most reactivity against GSH, but still looked clean across the enzyme panel. 2-bromodihydroisoxazoles showed a bit of reactivity, especially against MurE (a member of the panel), but no covalent binding could be confirmed by MS (must be reversible). Cyanoacetamides showed no reactivity at all, and neither did acyl imidazoles.

Now, there are electrophiles that are hot enough to cause trouble. You shouldn't expect clean behavior from an acid chloride or something, but the limits are well above where most of us think they are. If some of these compounds (like the imidazolides) had been profiled across an entire proteome, then perhaps something would have turned up at a low level (as Cravatt and Weerapana saw in that link in the first paragraph). But these things will vary compound by compound - some of them will find a place where they can sit long enough for a reaction to happen, and some of them won't. Here's what the authors conclude:

An unexpected but significant consequence of the present study is the relatively low inhibitory potential of the reactive compounds against the analyzed enzymes. Even in cytotoxicity assays and when we looked for inhibitor enzyme adduct formation we did not find any elevated cytotoxicity or unspecific modification of proteins. Particularly in the case of chloroacetylamides/-anilides and dimethylsulfonium salts, which we consider to be among the most reactive in this series, this is a promising result. From these results the following consequences for moderately reactive groups in medicinal chemistry can be drawn. Promiscuous reactivity and off-target effects of electrophiles with moderate reactivity may often be overestimated. It also does not appear justified to generally exclude “reactive” moieties from compound libraries for screening purposes, since the nonspecific reactivity may turn out to be much inferior than anticipated.

There are a lot of potentially useful compounds that most of us have never thought of looking at, because of our own fears. We should go there.

Comments (12) + TrackBacks (0) | Category: Chemical Biology | Chemical News | Drug Assays

September 5, 2014

A VC Firm Touts Its Successes - And Its Failures

Email This Entry

Posted by Derek

Here's what sounds like a good idea from VC firm Index Ventures, from the latest issue of BioCentury (same one I referenced the other day). Like many others in the biopharma venture capital world, they're trying to run the "killer experiment" as soon as possible, to see which ideas for new companies look solid. Unlike the others, though, they're planning a web site where they will detail the successes - and the failures. Here's an example:

Founded in 2013, (Purple Pharmaceuticals) was started to identify small molecule inhibitors of proprotein convertase subtilisin/kexin type 9. Two mAbs against PCSK9 are already in Phase III testing to treat hypercholesterolemia with regulatory submissions expected this year: evolocumab from Amgen Inc. and alirocumab from partners Regeneron Pharmaceuticals Inc. and Sanofi.

Grainger said the antibodies have limitations, as they require high doses to suppress PCSK9 activity and once-weekly or once-monthly infusions. Thus a pill that could match the PCSK9 inhibition of the biologics could be “the holy grail” of lowering LDL cholesterol.

Purple began by trying to identify small molecules that were highly selective for PCSK9 over other proprotein convertases because, as Grainger noted, “PCSK9 is a member of a large family of enzymes that do some pretty critical things.”

The killer experiment, he said, “was to ask could we make a small molecule that was selective over these other proprotein convertases, and could we demonstrate that it would lower LDL cholesterol?”

After a year, Purple had identified some hits selective for PCSK9, but a conversation with researchers at the Genentech Inc. unit of Roche led to the realization that the virtual company would need to run a second experiment.

“We learned from that interaction with Genentech that they had also run a PCSK9 screening program the same way we had,” Grainger said. “They discovered that their hits, while preventing PCSK9 from cleaving an external substrate, did not prevent PCSK9 from cleaving itself.”

Purple learned that in vivo, PCSK9 is auto-activated by cleaving itself — meaning the only important interaction to inhibit is PCSK9 auto- activation, not interactions with external substrates.

Purple’s second experiment showed none of its small molecule hits that inhibited PCSK9 interaction with an external substrate also inhibited auto-activation.

“Therefore we were able to kill a project which had spent at that stage only about £300,000 over a year, only to discover at the critical moment that it didn’t have the profile that we wanted,” Grainger said. “We were able to terminate that without having created any infrastructure, without having spent a painful amount of money prosecuting the project.”

That story illustrates a number of points about drug discovery. First off, congratulations to those involved for being able to definitively test a hypothesis; that's the engine at the heart of all scientific research. And as they say, it was good to be able to do that without having spent too much money and time, because both of those have a way of getting a bit out of control as complication after complication gets uncovered. Investors start getting jumpy when you keep coming back to them saying "Well, you know, it turns out that. . . ", but you know, it often turns out that way.

The next thing this story shows is that when you see an obvious gap in the landscape that there may well be a good reason for it. PCSK9 antibodies are widely thought to be potential blockbusters; a huge battle is shaping up in that area. So why no small molecules, eh? That's the question that launched Purple, it seems, and it's a valid one. But it turns out to have a valid answer, one that others in the field had already discovered. I suspect that the people behind this effort were, at the same time they were characterizing their lead molecules, also beating all the bushes for the sort of information that they obtained from Genentech. Somebody must have tried small-molecule PCSK9 inhibitors, you'd think, so what happened? Were those projects abandoned for good reasons, or was there still some opportunity there that a new company could claim for itself?

There may well be more to this story, though, than the Index Ventures people are saying. Update: there is - see the end of this post. The autocatalytic cleavage of PCSK9 was already well-known - pretty much everything in the that protease family works that way. (The difference is that with PCSK9, the prodomain part of the protein stays on longer - details of its cleavage were worked out in 2006). And in this 2008 paper from Journal of Lipid Research, we find this:

Several approaches for inhibiting PCSK9 function are theoretically feasible. Because autocatalytic cleavage is required for the maturation of PCSK9, a small-molecule inhibitor of autocatalysis might be useful, provided that it was specific for PCSK9 processing and did not lead to a toxic accumulation of misfolded PCSK9. Small molecules that block the PCSK9-LDL receptor interactions would likely be efficacious, although designing inhibitors of protein-protein interactions is a tall order. Antisense approaches pioneered by Isis Pharmaceuticals (Carlsbad, CA) are well suited for liver targets, and studies in mice suggest that this approach is efficacious for PCSK9. Finally, there is considerable interest in developing antibody therapeutics to inhibit PCSK9-LDL receptor interactions.

Even more to the point is the paper that that JLR piece is commenting on. That one demonstrates, through studies of mutated PCSK9 proteins, that its catalytic activity does not seem to be necessary at all for its effects on LDL receptors (a result that had already been suggested in cell assays). Taken together, you'd come away with the strong impression that inhibiting PCSK9's catalytic activity, other than stopping it from turning itself into its active form, had a low probability of doing anything to cholesterol levels. And you'd come away with that impression in 2008, at the latest.

So Purple's idea was a longer shot than it appeared on the surface, not that the real information was exactly buried deep in the literature. They shouldn't have needed someone at Genentech to tell them that PCSK9's autocatalysis was the real target - I've never worked in the area at all, and I found this out in fifteen minutes on PubMed while riding in to work. They must have had more reason to think that an assay for PCSK9's exogenous activity would be worth running - either that, or this story has gotten garbled along the way.

But this example aside, I applaud the idea of making these early-stage calls public. And I agree with the Index Ventures folks that this should actually help academics and others unused to drug discovery to see what needs to be done to actually launch an idea out into the world. I look forward to seeing the web site - and perhaps to hearing a bit more about what really happened at Purple.

Update: David Grainger of Index Ventures has more in the comments, and says that there is indeed more to the story. He points out that mutations of PCSK9 were found that inhibited its autocatalytic activity (such as this one), and that work had appeared that suggested that molecules that inhibited only the autocatalytic activity could be useful. This is what Purple was seeking - the BioCentury piece makes things sound a bit different (see above), but the problem seems to have been that molecules that inhibited PCSK9's activity against other substrates turned out not to inhibit its activity against itself. If I'm interpreting this right, then, Genentech's contribution was to point out that the autocatalytic activity couldn't be modeled by looking at another substrate.

Comments (5) + TrackBacks (0) | Category: Business and Markets | Cardiovascular Disease | Drug Assays

September 4, 2014

Are Your Compounds Ugly? Do You Know?

Email This Entry

Posted by Derek

A reader sends along this paper, on some small molecules targeting the C2 domain of coagulation factor VIII. It illustrates some points that have come up around here over the years, that's for sure. The target is not a particularly easy one: a hit would have to block the interaction of that protein domain with a membrane surface. There is something of a binding pocket down in that region, though, and there were some hits reported from a screen back in 2004. Overall, it looks like a lot of targets that show up, especially these days - you're trying to affect protein conformation by going after a not-necessarily-made-for-small-molecules cavity. Possible, but not something that's going to light up a screening deck, either.

And many of the things that do show up are going to be false positives of one sort or another. That's always the tricky part of doing low-hit-rate screening. The odds are excellent that any given "hit" will turn out not to be real, since the odds are against having any hits at all. This is especially a red flag when you screen something like this and you get a surprisingly decent hit rate. You should suspect fluorescence interference, aggregation, impurities, any of the other myriad ways that things can be troublesome rather than assume that gosh, this target is easier than we thought.

It's often a chemist who's in charge of dumping these buckets of cold water (if you have the help of the people who set up the assay, so much the better). Traditionally, it's one of the biology project champions who gets enthusiastic about the great list of compounds, but if you have someone who's been burned by false positives a few times, then so much the better, too. It's not fun to knock down all these "hits" and "leads", but someone's got to do it, otherwise everyone's time will be wasted to an even more painful extent.

And you should be especially worried when your screen turns up compounds like some of the ones in this paper. Yep, it's our old friends the rhodanines, everybody's cheap date of the screening deck. These compounds have come up around here many times, because they keep on showing up in the flippin' literature. In this case, the authors did some virtual screening over the ChemBridge collection and then moved on to assays against the protein itself, eventually finding a number of active compounds in the micromolar range. The compounds look a lot like the ones from 2004, since those were used as the template for screening, and that was a pretty ugly rhodanine-infested set, too.

Indeed most of the compounds they found are pretty unattractive - the aforementioned rhodanines, lots of nitroaromatics, some other heterocycles that also hit more often than one would like. I would feel better about these sorts of papers if the authors acknowledged somewhere that some of their structures are frequent hitters and might be problematic, but you don't often see that: a hit is a hit, and everything's equally valid, apparently. I would also feel better if there were something in the experimental section about how all the compounds were assayed by LC/MS and NMR, but you don't often see that, either, and I don't see it here. Implicitly trusting the label is not a good policy. Even if the particular compounds are the right ones in this case, not checking them shows a lack of experience (and perhaps too trusting a nature where organic chemistry is concerned).

But let's cross our fingers and assume that these are indeed the right compounds. What does it mean when your screening provides you with a bunch of structures like this? The first thing you can say is that your target is indeed a low-probability one for small molecules to bind to - if most everything you get is a promiscuous-looking ugly, then the suspicion is that only the most obliging compounds in a typical screening collection will bother looking at your binding site at all. And that means that if you want something better, you're really going to have to dig for it (and dig through a mound of false positives and still more frequent hitters to find it).

Why would you want to do that? Aren't these tool compounds, useful to find out more about the biology and behavior of the target? Well, that's the problem. If your compounds are rhodanines, or from other such badly-behaved classes, then they are almost completely unsuitable as tool compounds. You especially don't want to trust anything they're telling you in a cellular (or worse, whole-animal) assay, because there is just no telling what else they're binding to. Any readout from such an assay has to be viewed with great suspicion, and what kind of a tool is that?

Well then, aren't these starting points for further optimization? It's tempting to think so, and you can give it a try. But likely as not, the objectionable features are the ones that you can't get rid of very easily. If you could ditch those without paying too much of a penalty, you would have presumably found more appealing molecules in your original screen and skipped this stage altogether. You might be better off running a different sort of screen and trying for something outside of these classes, rather than trying to synthesize a silk purse out of said sow's ear. If you do start from such a structure, prepare for a lot of work.

As mentioned, the problem with a lot of papers that advance such structures is that they don't seem to be aware of these issues at all. If they are, they certainly don't being them up (which is arguably even worse). Then someone else comes along, who hasn't had a chance to learn any of this yet, either, and reads the paper without coming out any smarter. They may, in fact, have been made slightly less competent by reading it, because now they think that there are these good hits for Target Z, for one thing, and that the structures shown in the paper must be OK, because here they are in this paper, with no mention of any potential problems.

The problem is, there are a lot of interesting targets out there that tend to yield just these sorts of hits. My own opinion is that you can then say that yes, this target can (possibly) bind a small molecule, if those hits are in fact real, but just barely. If you don't even pick up any frequent hitters, you're in an even tougher bind, but if all you pick up are frequent hitters, it doesn't mean that things are that much easier.

Comments (21) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | The Scientific Literature

August 26, 2014

A New Look at Phenotypic Screening

Email This Entry

Posted by Derek

There have been several analyses that have suggested that phenotypic drug discovery was unusually effective in delivering "first in class" drugs. Now comes a reworking of that question, and these authors (Jörg Eder, Richard Sedrani, and Christian Wiesmann of Novartis) find plenty of room to question that conclusion.

What they've done is to deliberately focus on the first-in-class drug approvals from 1999 to 2013, and take a detailed look at their origins. There have been 113 such drugs, and they find that 78 of them (45 small molecules and 33 biologics) come from target-based approaches, and 35 from "systems-based" approaches. They further divide the latter into "chemocentric" discovery, based around known pharmacophores, and so on, versus pure from-the-ground-up phenotypic screening, and the 33 systems compounds then split out 25 to 8.

As you might expect, a lot of these conclusions depend on what you classify as "phenotypic". The earlier paper stopped at the target-based/not target-based distinction, but this one is more strict: phenotypic screening is the evaluation of a large number of compounds (likely a random assortment) against a biological system, where you look for a desired phenotype without knowing what the target might be. And that's why this paper comes up with the term "chemocentric drug discovery", to encompass isolation of natural products, modification of known active structures, and so on.

Such conclusions also depend on knowing what approach was used in the original screening, and as everyone who's written about these things admits, this isn't always public information. The many readers of this site who've seen a drug project go from start to finish will appreciate how hard it is to find an accurate retelling of any given effort. Stuff gets left out, forgotten, is un- (or over-)appreciated, swept under the rug, etc. (And besides, an absolutely faithful retelling, with every single wrong turn left in, would be pretty difficult to sit through, wouldn't it?) At any rate, by the time a drug reaches FDA approval, many of the people who were present at the project's birth have probably scattered to other organizations entirely, have retired or been retired against their will, and so on.

But against all these obstacles, the authors seem to have done as thorough a job as anyone could possibly do. So looking further at their numbers, here are some more detailed breakdowns. Of those 45 first-in-class small molecules, 21 were from screening (18 of those high-throughput screening, 1 fragment-based, 1 in silico, and one low-throughput/directed screening). 18 came from chemocentric approaches, and 6 from modeling off of a known compound.

Of the 33 systems-based drugs, those 8 that were "pure phenotypic" feature one antibody (alemtuzumab) which was raised without knowledge of its target, and seven small molecules: sirolimus, fingolimod, eribulin, daptomycin, artemether–lumefantrine, bedaquiline and trametinib. The first three of those are natural products, or derived from natural products. Outside of fingolimod, all of them are anti-infectives or antiproliferatives, which I'd bet reflects the comparative ease of running pure phenotypic assays with those readouts.

Here are the authors on the discrepancies between their paper and the earlier one:

At first glance, the results of our analysis appear to sig­nificantly deviate from the numbers previously pub­lished for first­-in­-class drugs, which reported that of the 75 first-­in-­class drugs discovered between 1999 and 2008, 28 (37%) were discovered through phenotypic screening, 17 (23%) through target-­based approaches, 25 (33%) were biologics and five (7%) came from other approaches. This discrepancy occurs for two reasons. First, we consider biologics to be target­-based drugs, as there is little philosophical distinction in the hypothesis­ driven approach to drug discovery for small­-molecule drugs versus biologics. Second, the past 5 years of our analysis time frame have seen a significant increase in the approval of first-­in-­class drugs, most of which were discovered in a target­-based fashion.

Fair enough, and it may well be that many of us have been too optimistic about the evidence for the straight phenotypic approach. But the figure we don't have (and aren't going to get) is the overall success rate for both techniques. The number of target-based and phenotypic-based screening efforts that have been quietly abandoned - that's what we'd need to have to know which one has the better delivery percentage. If 78/113 drugs, 69% of the first-in-class approvals from the last 25 years, have come from target-based approaches how does that compare with the total number of first-in-class drug projects? My own suspicion is that target-based drug discovery has accounted for more than 70% of the industry's efforts over that span, which would mean that systems-based approaches have been relatively over-performing. But there's no way to know this for sure, and I may just be coming up with something that I want to hear.

That might especially be true when you consider that there are many therapeutic areas where phenotypic screening basically impossible (Alzheimer's, anyone?) But there's a flip side to that argument: it means that there's no special phenotypic sauce that you can spread around, either. The fact that so many of those pure-phenotypic drugs are in areas with such clear cellular readouts is suggestive. Even if phenotypic screeningwere to have some statistical advantage, you can't just go around telling people to be "more phenotypic" and expect increased success, especially outside anti-infectives or antiproliferatives.

The authors have another interesting point to make. As part of their analysis of these 113 first-in-class drugs, they've tried to see what the timeline is from the first efforts in the area to an approved drug. That's not easy, and there are some arbitrary decisions to be made. One example they give is anti-angiogenesis. The first report of tumors being able to stimulate blood vessel growth was in 1945. The presence of soluble tumor-derived growth factors was confirmed in 1968. VEGF, the outstanding example of these, was purified in 1983, and was cloned in 1989. So when did the starting pistol fire for drug discovery in this area? The authors choose 1983, which seems reasonable, but it's a judgment call.

So with all that in mind, they find that the average lead time (from discovery to drug) for a target-based project is 20 years, and for a systems-based drug it's been 25 years. They suggest that since target-based drug discovery has only been around since the late 1980s or so, that its impact is only recently beginning to show up in the figures, and that it's in much better shape than some would suppose.

The data also suggest that target-­based drug discovery might have helped reduce the median time for drug discovery and development. Closer examination of the differences in median times between systems­-based approaches and target­-based approaches revealed that the 5-­year median difference in overall approval time is largely due to statistically significant differences in the period from patent publication to FDA approval, where target-­based approaches (taking 8 years) took only half the time as systems­-based approaches (taking 16 years). . .

The pharmaceutical industry has often been criticized for not being sufficiently innovative. We think that our analysis indicates otherwise and perhaps even suggests that the best is yet to come as, owing to the length of time between project initiation and launch, new technologies such as high­-throughput screening and the sequencing of the human genome may only be starting to have a major impact on drug approvals. . .

Now that's an optimistic point of view, I have to say. The genome certainly still has plenty of time to deliver, but you probably won't find too many other people saying in 2014 that HTS is only now starting to have an impact on drug approvals. My own take on this is that they're covering too wide a band of technologies with such statements, lumping together things that have come in at different times during this period and which would be expected to have differently-timed impacts on the rate of drug discovery. On the other hand, I would like this glass-half-full view to be correct, since it implies that things should be steadily improving in the business, and we could use it.

But the authors take pains to show, in the last part of their paper, that they're not putting down phenotypic drug discovery. In fact, they're calling for it to be strengthened as its own discipline, and not (as they put it) just as a falling back to the older "chemocentric" methods of the 1980s and before:

Perhaps we are in a phase today similar to the one in the mid­-1980s, when systems-­based chemocentric drug discovery was largely replaced by target­-based approaches. This allowed the field to greatly expand beyond the relatively limited number of scaffolds that had been studied for decades and to gain access to many more pharmacologically active compound classes, pro­viding a boost to innovation. Now, with an increased chemical space, the time might be right to further broaden the target space and open up new avenues. This could well be achieved by investing in phenotypic screening using the compound libraries that have been established in the context of target-­based approaches. We therefore consider phenotypic screening not as a neoclassical approach that reverts to a supposedly more successful systems­-based method of the past, but instead as a logical evolution of the current target­-based activi­ties in drug discovery. Moreover, phenotypic screening is not just dependent on the use of many tools that have been established for target-­based approaches; it also requires further technological advancements.

That seems to me to be right on target: we probably are in a period just like the mid-to-late 1980s. In that case, though, a promising new technology was taking over because it seemed to offer so much more. Today, it's more driven by disillusionment with the current methods - but that means, even more, that we have to dig in and come up with some new ones and make them work.

Comments (7) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

August 21, 2014

Fragonomics, Eh?

Email This Entry

Posted by Derek

Edward Zartler ("Teddy Z" of the Practical Fragments blog) has a short piece in the latest ACS Medicinal Chemistry Letters on fragment-based drug discovery. He applies the term "fragonomics" to the field (more on this in a moment), and provides a really useful overview of how it should work.

One of his big points is that fragment work isn't so much about using smaller-than-usual molecules, as it is using molecules that make only good interactions with the target.. It's just that smaller molecules are far more likely to achieve that - a larger one will have some really strong interactions, along with some things that actually hurt the binding. You can start with something large and hack pieces of it off, but that's often a difficult process (and you can't always recapitulate the binding mode, either). But if you have a smaller piece that only makes a positive interaction or two, then you can build out from that, tiptoeing around the various landmines as you go. That's the concept of "ligand efficiency", without using a single equation.

He also emphasizes that having a simpler molecule to work on means that the SAR can be tested and expanded quickly, often without anyone hitting the lab bench at all. You can order things up from the vendors or raid your own screening collection for close analogs. This delays the entry of the medicinal chemists to the project, which (considering that their time is always in demand) is a feature to be happy about.

The article ends up by saying that "Fragonomics has won the field. . .The age of the medchemist is over; now is the time of the biophysicist." I don't know if that's quite the way to win friends and influence people, though. Medicinal chemists are rather sensitive to threats to their existence (with good reason), so my worry is that coming on like this will make chemists who haven't tried it leery of fragment-based drug design in general. I'm also not thrilled with "fragonomics" as a term (just as I'm not thrilled with most of the newly-coined "omics" terms). The word doesn't add anything; it's just a replacement for having to say "fragment-based drug discovery" or "FBDD" all the time. It's not that we don't need a replacement for the unwieldy phrase - it's just that I think that many people might (by now) be ready to dismiss anything that's had "omics" slapped on it. I wish I had something better to offer, but I'm coming up blank myself.

Comments (39) + TrackBacks (0) | Category: Drug Assays

August 19, 2014

Don't Optimize Your Plasma Protein Binding

Email This Entry

Posted by Derek

Here's a very good review article in J. Med. Chem. on the topic of protein binding. For those outside the field, that's the phenomenon of drug compounds getting into the bloodstream and then sticking to one or more blood proteins. Human serum albumin (HSA) is a big player here - it's a very abundant blood protein that's practically honeycombed with binding sites - but there are several others. The authors (from Genentech) take on the disagreements about whether low plasma protein binding is a good property for drug development (and conversely, whether high protein binding is a warning flag). The short answer, according to the paper: neither one.

To further examine the trend of PPB for recently approved drugs, we compiled the available PPB data for drugs approved by the U.S. FDA from 2003 to 2013. Although the distribution pattern of PPB is similar to those of the previously marketed drugs, the recently approved drugs generally show even higher PPB than the previously marketed drugs (Figure 1). The PPB of 45% newly approved drugs is >95%, and the PPB of 24% is >99%. These data demonstrate that compounds with PPB > 99% can still be valuable drugs. Retrospectively, if we had posed an arbitrary cutoff value for the PPB in the drug discovery stage, we could have missed many valuable medicines in the past decade. We suggest that PPB is neither a good nor a bad property for a drug and should not be optimized in drug design.

That topic has come up around here a few times, as could be expected - it's a standard med-chem argument. And this isn't even the first time that a paper has come out warning people that trying to optimize on "free fraction" is a bad idea: see this 2010 one from Nature Reviews Drug Discovery.

But it's clearly worth repeating - there are a lot of people who get quite worked about about this number - in some cases, because they have funny-looking PK and are trying to explain it, or in some cases, just because it's a number and numbers are good, right?

Comments (15) + TrackBacks (0) | Category: Drug Assays | Drug Development | Pharmacokinetics

July 24, 2014

Phenotypic Assays in Cancer Drug Discovery

Email This Entry

Posted by Derek

The topic of phenotypic screening has come up around here many times, as indeed it comes up very often in drug discovery. Give your compounds to cells or to animals and look for the effect you want: what could be simpler? Well, a lot of things could, as anyone who's actually done this sort of screening will be glad to tell you, but done right, it's a very powerful technique.

It's also true that a huge amount of industrial effort is going into cancer drug discovery, so you'd think that there would be a natural overlap between these: see if your compounds kill or slow cancer cells, or tumors in an animal, and you're on track, right? But there's a huge disconnect here, and that's the subject of a new paper in Nature Reviews Drug Discovery. (Full disclosure: one of the authors is a former colleague, and I had a chance to look over the manuscript while it was being prepared). Here's the hard part:

Among the factors contributing to the growing interest in phenotypic screening in drug discovery in general is the perception that, by avoiding oversimplified reductionist assumptions regarding molecular targets and instead focusing on functional effects, compounds that are discovered in phenotypic assays may be more likely to show clinical efficacy. However, cancer presents a challenge to this perception as the cell-based models that are typically used in cancer drug discovery are poor surrogates of the actual disease. The definitive test of both target hypotheses and phenotypic models can only be carried out in the clinic. The challenge of cancer drug discovery is to maximize the probability that drugs discovered by either biochemical or phenotypic methods will translate into clinical efficacy and improved disease control.

Good models in living systems, which are vital to any phenotypic drug discovery effort, are very much lacking in oncology. It's not that you can't get plenty of cancer cells to grow in a dish - they'll take over your other cell cultures if they get a chance. But those aren't the cells that you're going to be dealing with in vivo, not any more. Cancer cells tend to be genetically unstable, constantly throwing off mutations, and the in vitro lines are adapted to living in cull culture. That's true even if you implant them back into immune-compromised mice (the xenograft models). The number of drugs that look great in xenograft models and failed out in the real world is too large to count.

So doing pure phenotypic drug discovery against cancer is very difficult - you go down a lot of blind alleys, which is what phenotypic screening is supposed to prevent. The explosion of knowledge about cellular pathways in tumor cells has led to uncountable numbers of target-driven approaches instead, but (as everyone has had a chance to find out), it's rare to find a real-world cancer patient who can be helped by a single-target drug. Gleevec is the example that everyone thinks of, but the cruel truth is that it's the exceptional exception. All those newspaper articles ten years ago that heralded a wonderful era of targeted wonder drugs for cancer? They were wrong.

So what to do? This paper suggests that the answer is a hybrid approach:

For the purpose of this article, we consider ‘pure’ phenotypic screening to be a discovery process that identifies chemical entities that have desirable biological (phenotypic) effects on cells or organisms without having prior knowledge of their biochemical activity or mode of action against a specific molecular target or targets. However, in practice, many phenotypically driven discovery projects are not target-agnostic; conversely, effective target-based discovery relies heavily on phenotypic assays. Determining the causal relationships between target inhibition and phenotypic effects may well open up new and unexpected avenues of cancer biology.

In light of these considerations, we propose that in practice a considerable proportion of cancer drug discovery falls between pure PDD and TDD, in a category that we term ‘mechanism-informed phenotypic drug discovery’ (MIPDD). This category includes inhibitors of known or hypothesized molecular targets that are identified and/or optimized by assessing their effects on a therapeutically relevant phenotype, as well as drug candidates that are identified by their effect on a mechanistically defined phenotype or phenotypic marker and subsequently optimized for a specific target-engagement MOA.

I've heard these referred to as "directed phenotypic screens", and while challenging, it can be a very fruitful way to go. Balancing the two ways of working is the tricky part: you don't want to slack up on the model just so it'll give you results, if those results aren't going to be meaningful. And you don't want to be so dogmatic about your target ideas that you walk away from something that could be useful, but doesn't fit your scheme. If you can keep all these factors in line, you're a real drug discovery scientist, and no mistake.

How hard this is can be seen from the paper's Table 1, where they look over the oncology approvals since 1999, and classify them by what approaches were used for lead discovery and lead optimization. There's a pile of 21 kinase inhibitors (and eight other compounds) over in the box where both phases were driven by inhibition of a known target. And there are ten compounds whose origins were in straight phenotypic screening, with various paths forward after that. But the "mechanism-informed phenotypic screen" category is the shortest list of the three lead discovery approaches: seven compounds, optimized in various ways. (The authors are upfront about the difficulties of assembling this sort of overview - it can be hard to say just what really happened during discovery and development, and we don't have the data on the failures).

Of those 29 pure-target-based drugs, 18 were follow-ons to mechanisms that had already been developed. At this point, you'd expect to hear that the phenotypic assays, by contrast, delivered a lot more new mechanisms. But this isn't the case: 14 follow-ons versus five first-in-class. This really isn't what phenotypic screening is supposed to deliver (and has delivered in the past), and I agree with the paper that this shows how difficult it has been to do real phenotypic discovery in this field. The few assays that translate to the clinic tend to keep discovering the same sorts of things. (And once again, the analogy to antibacterials comes to mind, because that's exactly what happens if you do a straight phenotypic screen for antibacterials. You find the same old stuff. That field, too, has been moving toward hybrid target/phenotypic approaches).

The situation might be changing a bit. If you look at the drugs in the clinic (Phase II and Phase III), as opposed to the older ones that have made it all the way through, there are still a vast pile of target-driven ones (mostly kinase inhibitors). But you can find more examples of phenotypic candidates, and among them an unusually high proportion of outright no-mechanism-known compounds. Those are tricky to develop in this field:

In cases where the efficacy arises from the engagement of a cryptic target (or mechanism) other than the nominally identified one, there is potential for substan- tial downside. One of the driving rationales of targeted discovery in cancer is that patients can be selected by pre- dictive biomarkers. Therefore, if the nominal target is not responsible for the actions of the drug, an incorrect diagnostic hypothesis may result in the selection of patients who will — at best — not derive benefit. For example, multiple clinical trials of the nominal RAF inhibitor sorafenib in melanoma showed no benefit, regardless of the BRAF mutation status. This is consistent with the evidence that the primary target and pharmacodynamic driver of efficacy for sorafenib is actually VEGFR2. The more recent clinical success of the bona fide BRAF inhibitor vemurafenib in melanoma demonstrates that the target hypothesis of BRAF for melanoma was valid.

So, if you're going to do this mechanism-informed phenotypic screening, just how do you go about it? High-content screening techniques are one approach: get as much data as possible about the effects of your compounds, both at the molecular and cellular level (the latter by imaging). Using better cell assays is crucial: make them as realistic as you can (three-dimensional culture, co-culture with other cell types, etc.), and go for cells that are as close to primary tissue as possible. None of this is easy, or cheap, but the engineer's triangle is always in effect ("Fast, Cheap, Good: Pick Any Two").

Comments (23) + TrackBacks (0) | Category: Cancer | Drug Assays | Drug Development

July 22, 2014

Put Them in Cells and Find Out

Email This Entry

Posted by Derek

So, when you put some diverse small molecules into cellular assays, how many proteins are they really hitting? You may know a primary target or two that they're likely to interact with, or (if you're doing phenotypic screening), you may not have any idea at all. But how many proteins (or other targets) are there that bind small molecules at all?

This is a question that many people are interested in, but hard data to answer it are not easily obtained. There have been theoretical estimates via several techniques, but (understandably) not too much experimental evidence. Now comes this paper from Ben Cravatt's group, and it's one of the best attempts yet.

What they've done is to produce a library of compounds, via Ugi chemistry, containing both a photoaffinity handle and an alkyne (for later "click" tagging). They'd done something similar before, but the photoaffinity group in that case was a benzophenone, which is rather hefty. This time they used a diazirine, which is both small and the precursor to a very reactive carbene once it's irradiated. (My impression is that the diazirine is the first thing to try if you're doing photoaffinity work, for just those reasons). They made a small set of fairly diverse compounds (about 60), with no particular structural biases in mind, and set out to see what these things would label.

They treated PC-3 cells (human prostate-cancer derived) with each member of the library at 10 µM, then hit them with UV to do the photoaffinity reaction, labeled with a fluorescent tag via the alkyne, and fished for proteins. What they found was a pretty wide variety, all right, but not in the nonselective shotgun style. Most compounds showed distinct patterns of protein labeling, and most proteins picked out distinct SAR from the compound set. They picked out six members of the library for close study, and found that these labeled about 24 proteins (one compound only picked up one target, while the most promiscuous compound labeled nine). What's really interesting is that only about half of these were known to have any small-molecule ligands at all. There were proteins from a number of different classes, and some (9 out of 24) weren't even enzymes, but rather scaffolding and signaling proteins (which wouldn't be expected to have many small-molecule binding possibilities).

A closer look at non-labeled versions of the probe compounds versus more highly purified proteins confirmed that the compounds really are binding as expected (in some cases, a bit better than the non-photoaffinity versions, in some cases worse). So even as small a probe as a diazirine is not silent, which is just what medicinal chemists would have anticipated. (Heck, even a single methyl or fluoro isn't always silent, and a good thing, too). But overall, what this study suggests is that most small molecules are going to hit a number of proteins (1 up to a dozen?) in any given cell with pretty good affinity. It also (encouragingly) suggests that there are more small-molecule binding sites than you'd think, with proteins that have not evolved for ligand responses still showing the ability to pick things up.

There was another interesting thing that turned up: while none of the Ugi compounds was a nonselective grab-everything compound, some of the proteins were. A subset of proteins tended to pick up a wide variety of the non-clickable probe compounds, and appear to be strong, promiscuous binders. Medicinal chemists already know a few of these things - CYP metabolizing enzymes, serum albumin, and so on. This post has some other suggestions. But there are plenty more of them out there, unguessable ones that we don't know about yet (in this case, PTGR and VDAC subtypes, along with NAMPT). There's a lot to find out.

Comments (7) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

July 9, 2014

Outsourced Assays, Now a Cause For Wonder?

Email This Entry

Posted by Derek

Here's a look at Emerald Biotherapeutics (a name that's unfortunately easy to confuse with several other former Emeralds in this space). They're engaged in their own drug research, but they also have lab services for sale, using a proprietary system that they say generates fast, reproducible assays.

On July 1 the company unveiled a service that lets other labs send it instructions for their experiments via the Web. Robots then complete the work. The idea is a variation on the cloud-computing model, in which companies rent computers by the hour from, Google, and Microsoft instead of buying and managing their own equipment. In this case, biotech startups could offload some of their basic tasks—counting cells one at a time or isolating proteins—freeing their researchers to work on more complex jobs and analyze results. To control the myriad lab machines, Emerald has developed its own computer language and management software. The company is charging clients $1 to $100 per experiment and has vowed to return results within a day.

The Bloomberg Businessweek piece profiling them does a reasonable job, but I can't tell if its author knows that there's already a good amount of outsourcing of this type already. Emerald's system does indeed sound fast, though. But rarely does the quickness of an assay turn out to be the real bottleneck in any drug discovery effort, so I'm not sure how much of a selling point that is. The harder parts are the ones that can't be automated: figuring out what sort of assay to run, and troubleshooting it so that it can be reliably run on high-throughput machines are not trivial processes, and they can take a lot of time and effort. Even more difficult is the step before any of that: figuring out what you're going to be assaying at all. What's your target? What are you screening for? What's the great idea behind the whole project? That stuff is never going to be automated at all, and it's the key to the whole game.

But when I read things like this, I wonder a bit:

While pursuing the antiviral therapy, Emerald began developing tools to work faster. Each piece of lab equipment, made by companies including Agilent Technologies (A) and Thermo Fisher Scientific (TMO), had its own often-rudimentary software. Emerald’s solution was to write management software that centralized control of all the machines, with consistent ways to specify what type of experiment to run, what order to mix the chemicals in, how long to heat something, and so on. “There are about 100 knobs you can turn with the software,” says Frezza. Crucially, Emerald can store all the information the machines collect in a single database, where scientists can analyze it. This was a major advance over the still common practice of pasting printed reports into lab notebooks.

Well, that may be common in some places, but in my own experience, that paste-the-printed-report stuff went out a long time ago. Talking up the ability to have all the assay data collected in one place sounds like something from about fifteen or twenty years ago, although the situation can be different for the small startups who would be using Emerald (or their competitors) for outsourced assay work. But I would still expect any CRO shop to provide something better than a bunch of paper printouts!

Emerald may well have something worth selling, and I wish them success with it. Reproducible assays with fast turnaround are always welcome. But this article's "Gosh everything's gone virtual now wow" take on it isn't quite in line with reality.

Comments (11) + TrackBacks (0) | Category: Drug Assays

June 25, 2014

Where's the Widest Variety of Chemical Matter?

Email This Entry

Posted by Derek

A look through some of the medicinal chemistry literature this morning got me to thinking: does anyone have any idea of which drug target has the most different/diverse chemical matter that's been reported against it? I realize that different scaffolds are in the eye of the beholder, so it's going to be impossible to come up with any exact counts. But I think that all the sulfonamides that hit carbonic anhydrase, for example, should for this purpose be lumped together: that interaction with the zinc is crucial, and everything else follows after. Non-sulfonamide CA inhibitors would each form a new class for each new zinc-interacting motif, and any compounds that don't hit the zinc at all (are there any?) would add to the list, too. Then you have allosteric compounds, which are necessarily going to look different than active-site inhibitors.

My guess is that some of the nuclear receptors would turn out to win this competition. They can have large, flexible binding pockets that seem to recognize a variety of chemotypes. So maybe this question should be divided up a bit more:

1. What enzyme is known to have the widest chemical variety of active-site inhibitors?

2. Which GPCR has the widest chemical variety of agonists? Antagonists? (The antagonists are going to win this one, surely).

3. And the the open field question asked above: what drug target of any kind has had the widest variety of molecules reported to act on it, in any fashion?

I don't imagine that we'll come to any definitive answer to any of these, but some people may have interesting nominations.

Update: in response to a query in the comments, maybe we should exempt the drug-metabolizing enzymes from the competition, since their whole reason for living is to take on a wide variety of unknown chemical structures.

Comments (25) + TrackBacks (0) | Category: Chemical News | Drug Assays

June 19, 2014

Dark Biology And Small Molecules

Email This Entry

Posted by Derek

Here's a discussion I got into the other day, in which I expressed some forceful opinions. I wanted to run it past a wider audience to see if I'm grounded in reality, or out on my own island (which has happened before).

Without getting into any details, we were talking about an area of potential drug research that has to do with transcriptional regulation. This one is clearly complicated - what part of transcription isn't complicated? But it's known that you can get things to happen by using things like epigenetic tool compounds (bromodomains, HDAC inhibitors, methyltransferases), and nuclear receptor ligands. None of these give you everything you want to see, by any means, but you do see some effect.

My take on this was that an effort to follow up with more epigenetic compounds and nuclear receptor ligands might well be a case of the classic "looking under the lamp-post because that's where the light is" syndrome. We don't have many small-molecule handles for affecting transcription, went my reasoning, and although such things are bromodomains, HDAC inhibition, and nuclear receptor signaling are wide-ranging, there's a lot more than the compounds in these spaces surely don't cover. In fact, given the wide range of these mechanisms, seeing a little tickling of any given transcriptional mechanism is about what I would expect from almost any of them, applied to almost anything. But that, to my mind, didn't necessarily mean that it was a lead worth following up.

My recommendation was for a phenotypic screen, if a good one could be worked up. There must be plenty of stuff going on with this system that we don't have any idea about, went my thinking. In the same way that the matter we can see through a telescope is only a tiny fraction of what appears to be really out there in the universe, I think that there's a vast amount of "dark biology" that we don't know much of. And the overwhelming majority of it has to be considered dark if we only consider the parts that we can light up with small molecules. For something that has to involve a huge array of protein-protein interactions, protein-nucleic acid interactions, and who knows what ancillary enzymes and binding sites, I wondered, what are the odds that the things that we happen know how to do with small molecules are the real answer?

So if you're going to dive into such waters (and many of you out there might be swimming around in them right now), by all means test whatever epigenetic and nuclear receptor compounds you might have around. Maybe you'll get a strong response. But if it all comes back as a little bit of this and a tiny bit of that, I'd say that these are unlikely to be convertible into robust drug mechanisms - the odds are that if there even is a robust drug mechanism out there, that you haven't hit it yet and that it will announce itself a bit more clearly if you manage to. A well-designed phenotypic screen might well be the best way to find such things, always keeping in mind that a badly designed phenotypic screen is the tar pit itself, the worst of both worlds.

So, am I too gloomy? Too jaded? Or simply a well-meaning realist? Thoughts welcome.

Comments (28) + TrackBacks (0) | Category: Drug Assays

June 13, 2014

Med-Chem, Automated?

Email This Entry

Posted by Derek

See what you think about this PDF: Cyclofluidics is advertising the "Robot Medicinal Chemist". It's an integrated microfluidics synthesis platform, assay/screening module, with software to decide what the next round of analogs should be:

Potential lead molecules are synthesised, purified and screened in fast serial mode, incorporating activity data from each compound as it is generated before selecting the next compound to make.

To ensure data quality, each compound is purified by integrated high pressure liquid chromatography (HPLC), its identity confirmed by mass spectrometry and the
concentration entering the assay determined in real time by evaporative light scattering detection (ELSD). The compound's IC50 is then measured in an on-line biochemical assay and this result fed into the design software before the algorithm selects the next compound to make – thus generating structure-activity relationship data. The system is designed to use interchangeable design algorithms, assay formats and chemistries and at any stage a medicinal chemist can intervene in order to adjust the design strategy.

I can see where this might work, but only in special cases. The chemistry part would seem to require a "core with substituents" approach, where a common intermediate gets various things hung off of it. (That's how a lot of medicinal chemistry gets done anyway). Flow chemistry has improved to where many reactions would be possible, but each new reaction type would have to be optimized a bit before you turned the machine loose, I'd think.

The assay part is more problematic. There are assays suitable for in-line evaluation like this, but there are plenty of others that aren't. (I would think that SPR would be particularly well-suited, since it operates in flow, anyway). But that prior optimization that the chemistry needs is needed even more here, to make sure that things are robust enough that the machine doesn't generate crappy numbers (and more swiftly than you could do by hand!)

The software is the part I'm really wondering about. How is this thing picking the next round of analogs? Physiochemical descriptions like logD? Some sort of expert system with med-chem lore in it? Does it do any modeling or conformational analysis? Inquiring minds want to know. And I'd also like to know if they've sold any of these systems so far, and to hear some comments from their users.

Update: here's one.

Comments (25) + TrackBacks (0) | Category: Chemical News | Drug Assays

June 9, 2014

Hosed-Up X-Ray Structures: A Big Problem

Email This Entry

Posted by Derek

X-ray crystallography is great stuff, no doubt about it. But it's not magic. It takes substantial human input to give a useful structure of a ligand bound to a protein - there are decisions to be made and differences to be split. It's important to emphasize, for those of us who are not crystallographers, that unless you have resolution down below 1Å - and I'll bet you don't - then your X-ray structures are not quite "structures"; they're models. A paper several years ago emphasized these factors for chemists outside the field.

About ten years ago, I wrote about this paper, which suggested that many ligand-bound structures seemed to have strain energy in them that wouldn't have been predicted. One interpretation is that there's more to ligand (and binding site) reorganization than people tend to realize, and that ligands don't always bind in their lowest-energy conformations. And while I still think that's true, the situation is complicated by another problem that's become more apparent over the years: many reported X-ray structures for ligand-bound proteins are just messed up.
Here's an editorial in ACS Medicinal Chemistry Letters that shows how bad the problem may well be. Reviews of the crystallographic databases have suggested that there are plenty of poorly refined structures hiding in there. But I didn't realize that they were as poorly refined as some of these. Take a look at the phosphate in 1xqd, and note how squashed-out those oxygens are around the first phosphorus. Or try the olefin in 4g93, which has been yanked 90 degrees out of plane. It's bad that there are such ridiculous structures in the literature, but the larger number of semi-plausible (but still wrong) structures is even worse.

Those structures at the left illustrate what's going on. The top one is an old PDB structure, 3qad, for an IKK inhibitor. It's a mess. Note that there's a tetrahedralish aromatic carbon (not happening), and a piperazine in a boat conformation (only slightly less unlikely). The structure was revised after this was pointed out to the middle version (3rzf), but that one still has some odd features - those two aromatic groups are flat-on in the same plane, and the amine between them and the next aryl is rather odd, too. Might be right, might be wrong - who's to know?

The most recent comprehensive look (from 2012) suggests that about 25% of the reported ligand-bound structures are mangled to the point of being misleading. This new editorial goes on to mention some computational tools that could help to keep this from happening, such as this one. If we're all going to draw conclusions from these things (and that's what they're there for, right?) we'd be better off using the best ones we can.

Comments (20) + TrackBacks (0) | Category: Drug Assays | In Silico

June 4, 2014

Predicting New Targets - Another Approach

Email This Entry

Posted by Derek

So you make a new chemical structure as part of a drug research program. What's it going to hit when it goes into an animal?

That question is a good indicator of the divide between the general public and actual chemists and pharmacologists. People without any med-chem background tend to think that we can predict these things, and people with it know that we can't predict much at all. Even just predicting activity at the actual desired target is no joke, and guessing what other targets a given compound might hit is, well, usually just guessing. We get surprised all the time.

That hasn't been for lack of trying, of course. Here's an effort from a few years ago on this exact question, and a team from Novartis has just published another approach. It builds on some earlier work of theirs (HTS fingerprints, HTSFP) that tries to classify compounds according to similar fingerprints of biological activity in suites of assays, rather than by their structures, and this latest one is called HTSFP-TID (target ID, and I think the acronym is getting a bit overloaded at that point).

We apply HTSFP-TID to make predictions for 1,357 natural products (NPs) and 1,416 experimental small molecules and marketed drugs (hereafter generally referred to as drugs). Our large-scale target prediction enables us to detect differences in the protein classes predicted for the two data sets, reveal target classes that so far have been underrepresented in target elucidation efforts, and devise strategies for a more effective targeting of the druggable genome. Our results show that even for highly investigated compounds such as marketed drugs, HTSFP-TID provides fresh hypotheses that were previously not pursued because they were not obvious based on the chemical structure of a molecule or against human intuition.

They have up to 230 or so assays to pick from, although it's for sure that none of the compounds have been through all of them. They required that any given compound have at least 50 different assays to its name, though (and these were dealt with as standard deviations off the mean, to keep things comparable). And what they found shows some interesting (and believable) discrepancies between the two sets of compounds. The natural product set gave mostly predictions for enzyme targets (70%), half of them being kinases. Proteases were about 15% of the target predictions, and only 4% were predicted GPCR targets. The drug-like set also predicted a lot of kinase interactions (44%), and this from a set where only 20% of the compounds were known to hit any kinases before. But it had only 5% protease target predictions, as opposed to 23% GPCR target predictions.

The group took a subset of compounds and ran them through new assays to see how the predictions came out, and the results weren't bad - overall, about 73% of the predictions were borne out by experiment. The kinase predictions, especially, seemed fairly accurate, although the GPCR calls were less so. They identified several new modes of action for existing compounds (a few of which they later discovered buried in the literature). They also tried a set of predictions based on chemical descriptor (the other standard approach), but found a lower hit rate. Interestingly, though, the two methods tended to give orthogonal predictions, which suggests that you might want to run things both ways if you care enough. Such efforts would seem particularly useful as you push into weirdo chemical or biological space, where we'll take whatever guidance we can get.

Novartis has 1.8 million compounds to work with, and plenty of assay data. It would be worth knowing what some other large collections would yield with the same algorithms: if you used (say) Merck's in-house data as a training set, and then applied it to all the compounds in the CHEMBL database, how similar would the set of predictions for them be? I'd very much like for someone to do something like this (and publish the results), but we'll see if that happens or not.

Comments (11) + TrackBacks (0) | Category: Drug Assays | In Silico

May 30, 2014

Covalent Fragments

Email This Entry

Posted by Derek

Many drug discovery researchers now have an idea of what to expect when a fragment library is screened against a new target. And some have had the experience of screening covalent, irreversible inhibitor structures against targets (a hot topic in recent years). But can you screen with a library of irreversibly-binding fragments?

This intersection has occurred to more than one group, but this paper marks the first published example that I know of. The authors, Alexander Statsyuk and co-workers at Northwestern, took what seems like a very sound approach. They were looking for compounds that would modify the active-site residues of cysteine proteases, which are the most likely targets in the proteome. But balancing the properties of a fragment collection with those of a covalent collection is tricky. Red-hot functional groups will certainly label your proteins, but they'll label the first things they see, which isn't too useful. If you go all the way in the other direction, epoxides are probably the least reactive covalent modifier, but they're so tame that unless they fit into a binding site perfectly, they might not do anything at all - and what are the chances that a fragment-sized molecule will bind that well? How much room is there in the middle?

That's what this paper is trying to find out. The team first surveyed a range of reactive functional groups against a test thiol, N-acetylcysteine. They attached an assortment of structures to each reactive end, and they were looking for two things: absolute reactivity of each covalent modifier, and how much it mattered as their structures varied. Acrylamides dropped out as a class because their more reactive examples were just too hot - their reactivity varied up to 2000x across a short range of examples. Vinylsulfonamides varied 8-fold, but acrylates and vinylsulfones were much less sensitive to structural variation. They picked acrylates as the less reactive of the two.

A small library of 100 diverse acrylates were then prepared (whose members still only varied about twofold in reactivity), and these were screened (100 micromolar) against papain as a prototype cysteine protease. They'd picked their fragments so that everything had a distinct molecular weight, so whole-protein mass spec could be used as a readout. Screening ten sets of ten mixtures showed that the enzyme picked out three distinct fragments from the entire set, a very encouraging result. Pretreatment of the enzyme with a known active-site labeling inhibitor shut down any reaction with the three hits, as it should have.

Keep in mind that this also means that 97 reasonably-sized acrylates were unable to label the very reactive Cys in the active site of papain, and that they did not label any surface residues. This suggests that the compounds that did make it in did so because of some structure-driven binding selectivity, which is just the territory that you want to be in. Adding an excess of glutathione to the labeling experiments did not shut things down, which also suggests that these are not-very-reactive acrylates whose structures are giving them an edge. Screen another enzyme, and you should pick up a different set of hits.

And that's exactly what they did next. Screening a rhinovirus cysteine protease (HRV3C) gave three totally new hits - not as powerful against that target as the other three were against papain, but real hits. Two other screens, against USP08 and UbcH7, did not yield any hits at all (except a couple of very weak ones against the former when the concentration was pushed hard). A larger reactive fragment library would seem to be the answer here; 100 compounds really isn't very much, even for fragment space, when you get down to it.

So this paper demonstrates that you can, in fact, find an overlap between fragment space and covalent inhibition, if you proceed carefully. Now here's a question that I'm not sure has ever been answered: if you find such a covalent fragment, and optimize it to be a much more potent binder, can you then pull the bait-and-switch by removing the covalent warhead, and still retain enough potency? Or is that too much to ask?

Comments (10) + TrackBacks (0) | Category: Chemical Biology | Chemical News | Drug Assays

May 12, 2014

DMSO Will Ruin Your Platinum Drugs. Take Heed.

Email This Entry

Posted by Derek

Here's the sort of experimental detail that can destroy a whole project if you're not aware of it. The platinum chemotherapy drugs are an odd class of things compared to the more typical organic compounds, but it's for sure that many of the people using them in research aren't aware of all of their peculiarities. One of those has been highlighted recently, and it's a sneaky one.

DMSO is, of course, the standard solvent used to take up test compounds for pharmacological assays. It's water-miscible and dissolves a huge range of organic compounds. Most of the time it's fine (unless you push its final concentration too high in the assay). But it's most definitely not fine for the platinum complexes. This paper shows that DMSO displaces the starting ligands, forming a new platinum complex that does not show the desired activity in cells. What's more, a look through the literature shows that up to one third of the reported in vitro studies on these compounds used DMSO to dissolve them, which throws their conclusions immediately into doubt. And since nearly half the papers did not even mention the solvent used, you'd have to think that DMSO showed up a good amount of the time in those as well.

What's even more disturbing is that these sorts of problems were first reported over twenty years ago, but it's clear that this knowledge has not made it into general circulation. So the word needs to get out: never dissolve cisplatin (or the related complexes) in DMSO, even though that might seem like the obvious thing to do. Editors and referees should take note as well.

Comments (14) + TrackBacks (0) | Category: Biological News | Drug Assays

April 10, 2014

Encoded Libraries Versus a Protein-Protein Interaction

Email This Entry

Posted by Derek

So here's the GSK paper on applying the DNA-encoded library technology to a protein-protein target. I'm particularly interested in seeing the more exotic techniques applied to hard targets like these, because it looks like there are plenty of them where we're going to need all the help we can get. In this case, they're going after integrin LFA-1. That's a key signaling molecule in leukocyte migration during inflammation, and there was an antibody (Raptiva, efalizumab) on the market, until it was withdrawn for too many side effects. (It dialed down the immune system rather too well). But can you replace an antibody with a small molecule?

A lot of people have tried. This is a pretty well-precedented protein-protein interaction for drug discovery, although (as this paper mentions), most of the screens have been direct PPI ones, and most of the compounds found have been allosteric - they fit into another spot on LFA-1 and disrupt the equilibrium between a low-affinity form and the high-affinity one. In this case, though, the GSK folks used their encoded libraries to screen directly against the LFA-1 protein. As usual, the theoretical number of compounds in the collection was bizarre, about 4 billion compounds (it's the substituted triazine library that they've described before).

An indanyl amino acid in one position on the triazine seemed to be a key SAR point in the resulting screen, and there were at least four other substituents at the next triazine point that kept up its activity. Synthesizing these off the DNA tags gave double-digit nanomolar affinities (if they hadn't, we wouldn't be hearing about this work, I'm pretty sure). Developing the SAR from these seems to have gone in classic med-chem fashion, although a lot of classic med-chem programs would very much like to be able to start off with some 50 nM compounds. The compounds were also potent in cell adhesion assays, with an interesting twist - the team also used a mutated form of LFA-1 where a disulfide holds it fixed in the high-affinity state. The known small-molecule allosteric inhibitors work against wild-type in this cell assay, but wipe out against the locked mutant, as they should. These triazines showed the same behavior; they also target the allosteric site.

That probably shouldn't have come as a surprise. Most protein-protein interactions have limited opportunities for small molecules to affect them, and if there's a known friendly spot like the allosteric site here, you'd have to expect that most of your hits are going to be landing on it. You wonder what might happen if you ran the ELT screen against the high-affinity-locked mutant protein - if it's good enough to work in cells, it should be good enough to serve in a screen for non-allosteric compounds. The answer (most likely) is that you sure wouldn't find any 50 nM leads - I wonder what you'd find at all? Running four billion compounds across a protein surface and finding no real hits would be a sobering experience.

The paper finishes up by showing the synthesis of some fluorescently tagged derivatives, and showing that these also work in cell assay. The last sentence is : "The latter phenomena provided an opportunity for ELT selections against a desired target in its natural state on cell surface. We are currently exploring this technology development opportunity." I wonder if they are? For the same reasons given above, you'd expect to find mostly allosteric binders, and those already seem to be findable. And it's my impression that this is the early-stage ELT stuff (the triazine library), plus, when you look at the list of authors, there are several "Present address" footnotes. So this work was presumably done a while back and is just now coming into the light.

So the question of using this technique against PPI targets remains open, as far as I can tell. This one had already been shown to yield small-molecule hits, and it did so again, in the same binding pocket. What happens when you set out into the unknown? Presumably, GlaxoSmithKline (and the other groups pursuing encoded libraries) know a lot more about than the rest of us do. Surely some screens like this have been run. Either they came up empty - in which case we'll never hear about them - or they actually yielded something interesting, in which case we'll hear about them over the next few years. If you want to know the answer before then, you're going to have to run some yourself. Isn't that always the way?

Comments (17) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

April 2, 2014

Binding Assays, Inside the Actual Cells

Email This Entry

Posted by Derek

Many readers will be familiar, at least in principle, with the "thermal shift assay". It goes by other names as well, but the principle is the same. The idea is that when a ligand binds to a protein, it stabilizes its structure to some degree. This gets measured by watching its behavior as samples of bound and unbound proteins are heated up, and the most common way to detect those changes in protein structure (and stability) is by using a fluorescent dye. Thus another common name for the assay, DSF, for Differential Scanning Fluorimetry. The dye has a better chance to bind to the newly denatured protein once the heat gets to that point, and that binding even can be detected by increasing fluorescence. The assay is popular, since it doesn't require much in specialized equipment and is pretty straightforward to set up, compared to something like SPR. Here's a nice slide presentation that's up on the web from UC Santa Cruz, and here's one of many articles on using the technique for screening.

I bring this up because of this paper last suumer in Science, detailing what the authors (a mixed team from Sweden and Singapore) called CETSA, the cellular thermal shift assay. They trying to do something that is very worthwhile indeed: measuring ligand binding inside living cells. Someone who's never done drug discovery might imagine that that's the sort of thing that we do all the time, but in reality, it's very tricky. You can measure ligand binding to an isolated protein in vitro any number of ways (although they may or may not give you the same answer!), and you can measure downstream effects that you can be more (or less) confident are the result of your compound binding to a cellular target. But direct binding measurements in a living cell are pretty uncommon.

I wish they weren't. Your protein of interest is going to be a different beast when it's on the job in its native environment, compared to sitting around in a well in some buffer solution. There are other proteins for it to interact with, a whole local environment that we don't know enough to replicate. There are modifications to its structure (phosphorylation and others) that you may or may not be aware of, which can change things around. And all of these have a temporal dimension, changing under different cellular states and stresses in ways that are usually flat-out impossible to replicate ex vivo.

Here's what this new paper proposes:

We have developed a process in which multiple aliquots of cell lysate were heated to different temperatures. After cooling, the samples were centrifuged to separate soluble fractions from precipitated proteins. We then quantified the presence of the target protein in the soluble fraction by Western blotting . . .

Surprisingly, when we evaluated the thermal melt curve of four different clinical drug targets in lysates from cultured mammalian cells, all target proteins showed distinct melting curves. When drugs known to bind to these proteins were added to the cell lysates, obvious shifts in the melting curves were detected. . .

That makes it sound like the experiments were all done after the cells were lysed, which wouldn't be that much of a difference from the existing thermal shift assays. But reading on, they then did this experiment with methotrexate and its enzyme target, dihydrofolate reductase (DHFR), along with ralitrexed and its target, thymidylate synthase:

DHFR and TS were used to determine whether CETSA could be used in intact cells as well as in lysates. Cells were exposed to either methotrexate or raltitrexed, washed, heated to different temperatures, cooled, and lysed. The cell lysates were cleared by centrifugation, and the levels of soluble target protein were measured, revealing large thermal shifts for DHFR and TS in treated cells as compared to controls. . .

So the thermal shift part of the experiment is being done inside the cells themselves, and the readout is the amount of non-denatured protein left after lysis and gel purification. That's ingenious, but it's also the sort of idea that (if it did occur to you) you might dismiss as "probably not going to work" and/or "has surely already been tried and didn't work". It's to this team's credit that they ran with it. This proves once again the soundness of Francis Crick's advice (in his memoir What Mad Pursuitand other places) to not pay too much attention to your own reasoning about how your ideas must be flawed. Run the experiment and see.

A number of interesting controls were run. Cell membranes seem to be intact during the heating process, to take care of one big worry. The effect of ralitrexed added to lysate was much greater than when it was added to intact cells, suggesting transport and cell penetration effects. A time course experiment showed that it took two to three hours to saturate the system with the drug. Running the same experiment on starved cells gave a lower effect, and all of these point towards the technique doing what it's supposed to be doing - measuring the effect of drug action in living cells under real-world conditions.

There's even an extension to whole animals, albeit with a covalent compound, the MetAP2 inhibitor TNP-470. It's a fumagillin derivative, so it's a diepoxide to start off, with an extra chloroacetamide for good measure. (You don't need that last reactive group, by the way, as Zafgen's MetAP2 compound demonstrates). The covalency gives you every chance to see the effect if it's going to be seen. Dosing mice with the compound, followed by organ harvesting, cell lysis, and heating after the lysis step showed that it was indeed detectable by thermal shift after isolation of the enzyme, in a dose-responsive manner, and that there was more of it in the kidneys than the liver.

Back in the regular assay, they show several examples of this working on other enzymes, but a particularly good one is PARP. Readers may recall the example of iniparib, which was taken into the clinic as a PARP-1 inhibitor, failed miserably, and was later shown not to really be hitting the target at all in actual cells and animals, as opposed to in vitro assays. CETSA experiments on it versus olaparib, which really does work via PARP-1, confirm this dramatically, and suggest that this assay could have told everyone a long time ago that there was something funny about iniparib in cells. (I should note that PARP has also been a testbed for other interesting cell assay techniques).

This leads to a few thoughts on larger questions. Sanofi went ahead with iniparib because it worked in their assays - turns out it just wasn't working through PARP inhibition, but probably by messing around with various cysteines. They were doing a phenotypic program without knowing it. This CETSA technique is, of course, completely target-directed, unless you feel like doing thermal shift measurements on a few hundred (or few thousand) proteins. But that makes me wonder if that's something that could be done. Is there some way to, say, impregnate the gel with the fluorescent shift dye and measure changes band by band? Probably not (the gel would melt, for one thing), but I (or someone) should listen to Francis Crick and try some variation on this.

I do have one worry. In my experience, thermal shift assays have not been all that useful. But I'm probably looking at a sampling bias, because (1) this technique is often used for screening fragments, where the potencies are not very impressive, and (2) it's often broken out to be used on tricky targets that no one can figure out how to assay any other way. Neither of those are conducive to seeing strong effects; if I'd been doing it on CDK4 or something, I might have a better opinion.

With that in mind, though, I find the whole CETSA idea very interesting, and well worth following up on. Time to look for a chance to try it out!

Comments (34) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

March 27, 2014

Another Target Validation Effort

Email This Entry

Posted by Derek

Here's another target validation initiative, with GSK, the EMBL, and the Sanger Institute joining forces. It's the Centre for Therapeutic Target Validation (CCTV):

CTTV scientists will combine their expertise to explore and interpret large volumes of data from genomics, proteomics, chemistry and disease biology. The new approach will complement existing methods of target validation, including analysis of published research on known biological processes, preclinical animal modelling and studying disease epidemiology. . .

This new collaboration draws on the diverse, specialised skills from scientific institutes and the pharmaceutical industry. Scientists from the Wellcome Trust Sanger Institute will contribute their unique understanding of the role of genetics in health and disease and EMBL-EBI, a global leader in the analysis and dissemination of biological data, will provide bioinformatics-led insights on the data and use its capabilities to integrate huge streams of different varieties of experimental data. GSK will contribute expertise in disease biology, translational medicine and drug discovery.

That's about as much detail as one could expect for now. It's hard to tell what sorts of targets they'll be working on, and by "what sorts" I mean what disease areas, what stage of knowledge, what provenance, and everything else. But the press release goes on to say that the information gathered by this effort will be open to the rest of the scientific community, which I applaud, and that should give us a chance to look under the hood a bit.

It's hard for me to say anything bad about such an effort, other than wishing it done on a larger scale. I was about to say "other than wishing it ten times larger", but I think I'd rather have nine other independent efforts set up than making this one huge, for several reasons. Quis validet ipsos validares, if that's a Latin verb and I haven't mangled it: Who will validate the validators? There's enough trickiness and uncertainty in this stuff for plenty more people to join in.

Comments (11) + TrackBacks (0) | Category: Biological News | Drug Assays

March 25, 2014

A New Way to Study Hepatotoxicity

Email This Entry

Posted by Derek

Every medicinal chemist fears and respects the liver. That's where our drugs go to die, or at least to be severely tested by that organ's array of powerful metabolizing enzymes. Getting a read on a drug candidate's hepatic stability is a crucial part of drug development, but there's an ever bigger prize out there: predicting outright liver toxicity. That, when it happens, is very bad news indeed, and can torpedo a clinical compound that seemed to be doing just fine - up until then.

Unfortunately, getting a handle on liver tox has been difficult, even with such strong motivation. It's a tough problem. And given that most drugs are not hepatotoxic, most of the time, any new assay that overpredicts liver tox might be even worse than no assay at all. There's a paper in the latest Nature Biotechnology, though, that looks promising.

What the authors (from Stanford and Toronto) are doing is trying to step back to the early mechanism of liver damage. One hypothesis has been that the production of reactive oxygen species (ROS) inside hepatic cells is the initial signal of trouble. ROS are known to damage biomolecules, of course. But more subtly, they're also known to be involved in a number of pathways used to sense that cellular damage (and in that capacity, seem to be key players in inducing the beneficial effects of exercise, among other things). Aerobic cells have had to deal with the downsides of oxygen for so long that they've learned to make the most of it.
This work (building on some previous studies from the same group) uses polymeric nanoparticles. They're semiconductors, and hooked up to be part of a fluorescence or chemiluminescence readout. (They use FRET for peroxynitrite and hypochlorite detection, more indicative of mitochondrial toxicity, and CRET for hydrogen peroxide, more indicative of Phase I metabolic toxicity). The particles are galactosylated to send them towards the liver cells in vivo, confirmed by necropsy and by confocal imaging. The assay system seemed to work well by itself, and in mouse serum, so they dosed it into mice and looked for what happened when the animals were given toxic doses of either acetominophen or isoniazid (both well-known hepatotox compounds at high levels). And it seems to work pretty well - they could image both the fluorescence and the chemiluminescence across a time course, and the dose/responses make sense. It looks like they're picking up nanomolar to micromolar levels of reactive species. They could also show the expected rescue of the acetominophen toxicity with some known agents (like GSH), but could also see differences between them, both in the magnitude of the effects and their time courses as well.

The chemiluminescent detection has been done before, as has the FRET one, but this one seems to be more convenient to dose, and having both ROS detection systems going at once is nice, too. One hopes that this sort of thing really can provide a way to get a solid in vivo read on hepatotoxicity, because we sure need one. Toxicologists tend to be a conservative bunch, with good reason, so don't look for this to revolutionize the field by the end of the year or anything. But there's a lot of promise here.

There are some things to look out for, though. For one, since these are necessarily being done in rodents, there will be differences in metabolism that will have to be taken into account, and some of those can be rather large. Not everything that injures a mouse liver will do so in humans, and vice versa. It's also worth remembering that hepatotoxicity is also a major problem with marketed drugs. That's going to be a much tougher problem to deal with, because some of these cases are due to overdose, some to drug-drug interactions, some to drug-alcohol interactions, and some to factors that no one's been able to pin down. One hopes, though, that if more drugs come through that show a clean liver profile that these problems might ameliorate a bit.

Comments (14) + TrackBacks (0) | Category: Drug Assays | Drug Development | Pharmacokinetics | Toxicology

March 20, 2014

Years Worth of the Stuff

Email This Entry

Posted by Derek

bAP15.pngThis time last year I mentioned a particularly disturbing-looking compound, sold commercially as a so-called "selective inhibitor" of two deubiquitinase enzymes. Now, I have a fairly open mind about chemical structures, but that thing is horrible, and if it's really selective for just those two proteins, then I'm off to truck-driving school just like Mom always wanted.

Here's an enlightening look through the literature at this whole class of compound, which has appeared again and again. The trail seems to go back to this 2001 paper in Biochemistry. By 2003, you see similar motifs showing up as putative anticancer agents in cell assays, and in 2006 the scaffold above makes its appearance in all its terrible glory.

The problem is, as Jonathan Baell points out in that post, that this series has apparently never really had a proper look at its SAR, or at its selectivity. It wanders through a series of publications full of on-again off-again cellular readouts, with a few tenuous conclusions drawn about its structure - and those are discarded or forgotten by the time the next paper comes around. As Baell puts it:

The dispiriting thing is that with or without critical analysis, this compound is almost certainly likely to end up with vendors as a “useful tool”, as they all do. Further, there will be dozens if not hundreds of papers out there where entirely analogous critical analyses of paper trails are possible.

The bottom line: people still don’t realize how easy it is to get a biological readout. The more subversive a compound, the more likely this is. True tools and most interesting compounds usually require a lot more medicinal chemistry and are often left behind or remain undiscovered.

Amen to that. There is way too much of this sort of thing in the med-chem literature already. I'm a big proponent of phenotypic screening, but setting up a good one is harder than setting up a good HTS, and working up the data from one is much harder than working up the data from an in vitro assay. The crazier or more reactive your "hit" seems to be, the more suspicious you should be.

The usual reply to that objection is "Tool compound!" But the standards for a tool compound, one used to investigate new biology and cellular pathways, are higher than usual. How are you going to unravel a biochemical puzzle if you're hitting nine different things, eight of which you're totally unaware of? Or skewing your assay readouts by some other effect entirely? This sort of thing happens all the time.

I can't help but think about such things when I read about a project like this one, where IBM's Watson software is going to be used to look at sequences from glioblastoma patients. That's going to be tough, but I think it's worth a look, and the Watson program seems to be just the correlation-searcher for the job. But the first thing they did was feed in piles of biochemical pathway data from the literature, and the problem is, a not insignificant proportion of that data is wrong. Statements like these are worrisome:

Over time, Watson will develop its own sense of what sources it looks at are consistently reliable. . .if the team decides to, it can start adding the full text of articles and branch out to other information sources. Between the known pathways and the scientific literature, however, IBM seems to think that Watson has a good grip on what typically goes on inside cells.

Maybe Watson can tell the rest of us, then. Because I don't know of anyone actually doing cell biology who feels that way, not if they're being honest with themselves. I wish the New York Genome Center and IBM luck in this, and I still think it's a worthwhile thing to at least try. But my guess is that it's going to be a humbling experience. Even if all the literature were correct in every detail, I think it would be one. And the literature is not correct in every detail. It has compounds like that one at the top of the entry in it, and people seem to think that they can draw conclusions from them.

Comments (19) + TrackBacks (0) | Category: Biological News | Cancer | Chemical Biology | Drug Assays | The Scientific Literature

March 18, 2014

Another DNA-Barcoded Program From GSK

Email This Entry

Posted by Derek

Two more papers have emerged from GSK using their DNA-encoded library platform. I'm always interested to see how this might be working out. One paper is on compounds for the tuberculosis target InhA, and the other is aimed at a lymphocyte protein-protein target, LFA-1. (I've written about this sort of thing previously here, here, and here).

Both of these have some interesting points - I'll cover the LFA-1 work in another post, though. InhA, for its part, is the target of the well-known tuberculosis drug isoniazid, and it has had (as you'd imagine) a good amount of attention over the years, especially since it's not the cleanest drug in the world (although it sure beats having tuberculosis). It's known to be a prodrug for the real active species, and there are also some nasty resistant strains out there, so there's certainly room for something better.
In this case, the GSK group apparently screened several of their DNA-encoded libraries against the target, but the paper only details what happened with one of them, the aminoproline scaffold shown. That would seem to be a pretty reasonable core, but it was one of 22 diamino acids in the library. R1 was 855 different reactants (amide formation, reductive amination, sulfonamides, ureas), and R2 was 857 of the same sorts of things, giving you, theoretically, a library of over 16 million compounds. (If you totaled up the number across the other DNA-encoded libraries, I wonder how many compounds this target saw in total?) Synthesizing a series of hits from this group off the DNA bar codes seems to have worked well, with one compound hitting in the tens of nanomolar range. (The success rate of this step is one of the things that those of us who haven't tried this technique are very interested in hearing about).
They even pulled out an InhA crystal structure with the compound shown, which really makes this one sound like a poster-child example of the whole technique (and might well be why we're reading about it in J. Med. Chem.) The main thing not to like about the structure is that it has three amides in it, but this is why one runs PK experiments, to see if having three amides is going to be a problem or not. A look at metabolic stability showed that it probably wasn't a bad starting point. Modifying those three regions gave them a glycine methyl ester at P1, which had better potency in both enzyme and cell assays. When you read through the paper, though, it appears that the team eventually had cause to regret having pursued it. A methyl ester is always under suspicion, and in this case it was justified: it wasn't stable under real-world conditions, and every attempt to modify it led to unacceptable losses in activity. It looks like they spent quite a bit of time trying to hang on to it, only to have to give up on it anyway.

In the end, the aminoproline in the middle was still intact (messing with it turned out to be a bad idea). The benzofuran was still there (nothing else was better). The pyrazole had extended from an N-methyl to an N-ethyl (nothing else was better there, either), and the P1 group was now a plain primary amide. A lot of med-chem programs work out like that - you go all around the barn and through the woods, emerging covered with mud and thorns only to find your best compound about fifteen feet away from where you started.

That compound, 65 in the paper, showed clean preliminary tox, along with good PK, potency, and selectivity. In vitro against the bacteria, it worked about as well as the fluoroquinolone moxifloxacin, which is a good level to hit. Unfortunately, when it was tried out in an actual mouse TB infection model, it did basically nothing at all. This, no doubt, is another reason that we're reading about this in J. Med. Chem.. When you read a paper from an industrial group in that journal, you're either visiting a museum or a mausoleum.

That final assay must have been a nasty moment for everyone, and you get the impression that there's still not an explanation for this major disconnect. It's hard to say if they saw it coming - had other compounds been in before, or did the team just save this assay for last and cross their fingers? But either way, the result isn't the fault of the DNA-encoded assay that provided the starting series - that, in this case, seems to have worked exactly as it was supposed to, and up to the infectious animal model study, everything looked pretty good.

Comments (24) + TrackBacks (0) | Category: Chemical Biology | Drug Assays | Infectious Diseases

February 24, 2014

Another Round of Stapled Peptide Wrangling

Email This Entry

Posted by Derek

When we last checked in on the Great Stapled Peptide Wars, researchers from Genentech, the Walter and Eliza Hall Institute and La Trobe University (the latter two in Australia) had questioned the usefulness and activity of several stapled Bim BH3 peptides. The original researchers (Walensky et al.) had then fired back strongly, pointing out that the criticisms seemed misdirected and directing the authors back to what they thought had been well-documented principles of working with such species.

Now the WEHI/Genentech/La Trobe group (Okamoto et al.) has responded, and it doesn't look like things are going to calm down any time soon. They'd made a lot of the 20-mer stapled peptide being inactive in cells, while the reply had been that yes, that's true, as you might have learned from reading the original papers again - it was the 21-mer that was active in cells. Okamoto and co-workers now say that they've confirmed this, but only in some cell lines - there are others for which the 21-mer is still inactive. What's more, they say that a modified but un-stapled 21-mer is just as active as the closed peptide, which suggests that the stapling might not be the key factor at all.

There's another glove thrown down (again). The earlier Genentech/WEHI/La Trobe paper had shown that the 20-mer had impaired binding to a range of Bcl target proteins. Walensky's team had replied that the 20-mer had been designed to have lower affinity, thus the poor binding results. But this new paper says that the 21-mer shows similarly poor binding behavior, so that can't be right, either.

This is a really short communication, and you get the impression that it was fired off as quickly as possible after the Walensky et al. rebuttal. There will, no doubt, be a reply. One aspect of it, I'm guessing, will be that contention about the unstapled peptide activity. I believe that the Walensky side of the argument have already shown that these substituted-but-unstapled peptides can show enhanced activity, probably due to cranking up their alpha-helical character (just not all the way to stapling them into that form). We shall see.

And this blowup reflects a lot of earlier dispute about Bcl, BAX/BAK peptides, and apoptosis in general. The WEHI group and others have been arguing out the details of these interactions in print for years, and this may be just another battlefield.

Comments (8) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

February 19, 2014

Ligand Efficiency: A Response to Shultz

Email This Entry

Posted by Derek

I'd like to throw a few more logs on the ligand efficiency fire. Chuck Reynolds of J&J (author of several papers on the subject, as aficionados know) left a comment to an earlier post that I think needs some wider exposure. I've added links to the references:

An article by Shultz was highlighted earlier in this blog and is mentioned again in this post on a recent review of Ligand Efficiency. Shultz’s criticism of LE, and indeed drug discovery “metrics” in general hinges on: (1) a discussion about the psychology of various metrics on scientists' thinking, (2) an assertion that the original definition of ligand efficiency, DeltaG/HA, is somehow flawed mathematically, and (3) counter examples where large ligands have been successfully brought to the clinic.

I will abstain from addressing the first point. With regard to the second, the argument that there is some mathematical rule that precludes dividing a logarithmic quantity by an integer is wrong. LE is simply a ratio of potency per atom. The fact that a log is involved in computing DeltaG, pKi, etc. is immaterial. He makes a more credible point that LE itself is on average non-linear with respect to large differences in HA count. But this is hardly a new observation, since exactly this trend has been discussed in detail by previous published studies (here, here, here, and here). It is, of course, true that if one goes to very low numbers of heavy atoms the classical definition of LE gets large, but as a practical matter medicinal chemists have little interest in extremely small fragments, and the mathematical catastrophe he warns us against only occurs when the number of heavy atoms goes to zero (with a zero in the denominator it makes no difference if there is a log in the numerator). Why would HA=0 ever be relevant to a med. chem. program? In any case a figure essentially equivalent to the prominently featured Figure 1a in the Shultz manuscript appears in all of the four papers listed above. You just need to know they exist.

With regard to the third argument, yes of course there are examples of drugs that defy one or more of the common guidelines (e.g MW). This seems to be a general problem of the community taking metrics and somehow turning them into “rules.” They are just helpful, hopefully, guideposts to be used as the situation and an organization’s appetite for risk dictate. One can only throw the concept of ligand efficiency out the window completely if you disagree with the general principle that it is better to design ligands where the atoms all, as much as possible, contribute to that molecule being a drug (e.g. potency, solubility, transport, tox, etc.). The fact that there are multiple LE schemes in the literature is just a natural consequence of ongoing efforts to refine, improve, and better apply a concept that most would agree is fundamental to successful drug discovery.

Well, as far as the math goes, dividing a log by an integer is not any sort of invalid operation. I believe that [log(x)]/y is the same as saying log(x to the one over y). That is, log(16) divided by 2 is the same as the log of 16 to the one-half power, or log(4). They both come out to about 0.602. Taking a BEI calculation as real chemistry example, a one-micromolar compound that weighs 250 would, by the usual definition, -log(Ki)/(MW/1000), have a BEI of 6/0.25, or 24. By the above rule, if you want to keep everything inside the log function, then say -log(0.0000001) divided by 0.25, that one-micromolar figure should be raised to the fourth power, then you take the log of the result (and flip the sign). One-millionth to the fourth power is one times ten to the minus twenty-fourth, so that gives you. . .24. No problem.

Shultz's objection that LE is not linear per heavy atom, though, is certainly valid, as Reynolds notes above as well. You have to realize this and bear it in mind while you're thinking about the topic. I think that one of the biggest problems with these metrics - and here's a point that both Reynolds and Shultz can agree on, I'll bet - is that they're tossed around too freely by people who would like to use them as a substitute for thought in the first place.

Comments (19) + TrackBacks (0) | Category: Drug Assays | Drug Development | In Silico

December 4, 2013

Cancer Cell Line Assays: You Won't Like Hearing This

Email This Entry

Posted by Derek

Here's some work that gets right to the heart of modern drug discovery: how are we supposed to deal with the variety of patients we're trying to treat? And the variety in the diseases themselves? And how does that correlate with our models of disease?

This new paper, a collaboration between eight institutions in the US and Europe, is itself a look at two other recent large efforts. One of these, the Cancer Genome Project, tested 138 anticancer drugs against 727 cell lines. Its authors said at the time (last year) that "By linking drug activity to the functional complexity of cancer genomes, systematic pharmacogenomic profiling in cancer cell lines provides a powerful biomarker discovery platform to guide rational cancer therapeutic strategies". The other study, the Cancer Cell Line Encyclopedia, tested 24 drugs against 1,036 cell lines. That one appeared at about the same time, and its authors said ". . .our results indicate that large, annotated cell-line collections may help to enable preclinical stratification schemata for anticancer agents. The generation of genetic predictions of drug response in the preclinical setting and their incorporation into cancer clinical trial design could speed the emergence of ‘personalized’ therapeutic regimens."

Well, will they? As the latest paper shows, the two earlier efforts overlap to the extent of 15 drugs, 471 cell lines, 64 genes and the expression of 12,153 genes. How well do they match up? Unfortunately, the answer is "Not too well at all". The discrepancies really come out in the drug sensitivity data. The authors tried controlling for all the variables they could think of - cell line origins, dosing protocols, assay readout technologies, methods of estimating IC50s (and/or AUCs), specific mechanistic pathways, and so on. Nothing really helped. The two studies were internally consistent, but their cross-correlation was relentlessly poor.

It gets worse. The authors tried the same sort of analysis on several drugs and cell lines themselves, and couldn't match their own data to either of the published studies. Their take on the situation:

Our analysis of these three large-scale pharmacogenomic studies points to a fundamental problem in assessment of pharmacological drug response. Although gene expression analysis has long been seen as a source of ‘noisy’ data, extensive work has led to standardized approaches to data collection and analysis and the development of robust platforms for measuring expression levels. This standardization has led to substantially higher quality, more reproducible expression data sets, and this is evident in the CCLE and CGP data where we found excellent correlation between expression profiles in cell lines profiled in both studies.

The poor correlation between drug response phenotypes is troubling and may represent a lack of standardization in experimental assays and data analysis methods. However, there may be other factors driving the discrepancy. As reported by the CGP, there was only a fair correlation (rs < 0.6) between camptothecin IC50 measurements generated at two sites using matched cell line collections and identical experimental protocols. Although this might lead to speculation that the cell lines could be the source of the observed phenotypic differences, this is highly unlikely as the gene expression profiles are well correlated between studies.

Although our analysis has been limited to common cell lines and drugs between studies, it is not unreasonable to assume that the measured pharmacogenomic response for other drugs and cell lines assayed are also questionable. Ultimately, the poor correlation in these published studies presents an obstacle to using the associated resources to build or validate predictive models of drug response. Because there is no clear concordance, predictive models of response developed using data from one study are almost guaranteed to fail when validated on data from another study, and there is no way with available data to determine which study is more accurate. This suggests that users of both data sets should be cautious in their interpretation of results derived from their analyses.

"Cautious" is one way to put it. These are the sorts of testing platforms that drug companies are using to sort out their early-stage compounds and projects, and very large amounts of time and money are riding on those decisions. What if they're gibberish? A number of warning sirens have gone off in the whole biomarker field over the last few years, and this one should be so loud that it can't be ignored. We have a lot of issues to sort out in our cell assays, and I'd advise anyone who thinks that their own data are totally solid to devote some serious thought to the possibility that they're wrong.

Here's a Nature News summary of the paper, if you don't have access. It notes that the authors of the two original studies don't necessarily agree that they conflict! I wonder if that's as much a psychological response as a statistical one. . .

Comments (21) + TrackBacks (0) | Category: Biological News | Cancer | Chemical Biology | Drug Assays

November 19, 2013

Phenotypic Screening Everywhere

Email This Entry

Posted by Derek

The Journal of Biomolecular Screening has a new issue devoted to phenotypic and functional screening approaches, and there looks to be some interesting material in there. The next issue will be Part II (they got so many manuscripts that the intended single issue ran over), and it all seems to have been triggered by the 2011 article in Nature Reviews Drug Discovery that I blogged about here. The Society for Laboratory Automation and Screening set up a special interest group for phenotypic drug discovery after that paper came out, and according to the lead editorial in this new issue, it quickly grew to become the largest SIG and one of the most active.
The reason for this might well be contained in the graphic shown, which is based on data from Bernard Munos. I'm hoping that those historical research spending numbers have been adjusted for inflation, but I believe that they have (since they were in Munos's original paper).

There's an update to the original Swinney and Anthony NRDD paper in this issue, too, and I'll highlight that in another post.

Comments (29) + TrackBacks (0) | Category: Drug Assays | Drug Industry History

October 29, 2013

Unraveling An Off-Rate

Email This Entry

Posted by Derek

Medicinal chemists talk a lot more about residence time and off rate than they used to. It's become clear that (at least in some cases) a key part of a drug's action is its kinetic behavior, specifically how quickly it leaves its binding site. You'd think that this would correlate well with its potency, but that's not necessarily so. Binding constants are a mix of on- and off-rates, and you can get to the same number by a variety of different means. Only if you're looking at very similar compounds with the same binding modes can you expect the correlation your intuition is telling you about, and even then you don't always get it.

There's a new paper in J. Med. Chem. from a team at Boehringer Ingelheim that takes a detailed look at this effect. The authors are working out the binding to the muscarinic receptor ligand tiotropium, which has been around a long time. (Boehringer's efforts in the muscarinic field have been around a long time, too, come to think of it). Tiotropium binds to the m2 subtype with a Ki of 0.2 nM, and to the m3 subtype with a Ki of 0.1 nM. But the compound has a much slower off rate on the m3 subtype, enough to make it physiologically distinct as an m3 ligand. Tiotropium is better known by its brand name Spiriva, and if its functional selectivity at the m3 receptors in the lungs wasn't pretty tight, it wouldn't be a drug. By carefully modifying its structure and introducing mutations into the receptor, this group hoped to figure out just why it's able to work the way it does.
The static details of tiotropium binding are well worked out - in fact, there's a recent X-ray structure, adding to the list of GPCRs that have been investigated by X-ray crystallography. There are plenty of interactions, as those binding constants would suggest:

The orthosteric binding sites of hM3R and hM2R are virtually identical. The positively charged headgroup of the antimuscarinic agent binds to (in the class of amine receptors highly conserved) Asp3.32 (D1483.32) and is surrounded by an aromatic cage consisting of Y1493.33, W5046.48, Y5076.51, Y5307.39, and Y5347.43. In addition to that, the aromatic substructures of the ligands dig into a hydrophobic region close to W2004.57 and the hydroxy groups, together with the ester groups, are bidentally interacting with N5086.52, forming close to optimal double hydrogen bonds. . .

The similarity of these binding sites was brought home to me personally when I was working on making selective antagonists of these myself. (If you want a real challenge, try differentiating m2 and m4). The authors point out, though, and crucially, that if you want to understand how different compounds bind to these receptors, the static pictures you get from X-ray structures are not enough. Homology modeling helps a good deal, but only if you take its results as indicators of dynamic processes, and not just swapping out residues in a framework.

Doing point-by-point single changes in both the tiotropium structure and the the receptor residues lets you use the kinetic data to your advantage. Such similar compounds should have similar modes of dissociation from the binding site. You can then compare off-rates to the binding constants, looking for the ones that deviate from the expected linear relationship. What they find is that the first event when tiotropium leaves the binding site is the opening of the aromatic cage mentioned above. Mutating any of these residues led to a big effect on the off-rate compared to the effect on the binding constant. Mutations further up along the tunnel leading to the binding site behaved in the same way: pretty much identical Ki values, but enhanced off-rates.

These observations, the paper says with commendable honesty, don't help the medicinal chemists all that much in designing compounds with better kinetics. You can imagine finding a compound that takes better advantage of this binding (maybe), but you can also imagine spending a lot of time trying to do that. The interaction with the asapragine at residue 508 is more useful from a drug design standpoint:

Our data provide evidence that the double hydrogen interaction of N5086.52 with tiotropium has a crucial influence on the off-rates beyond its influence on Ki. Mutation of N5086.52 to alanine accelerates the dissociation of tiotropium more than 1 order of magnitude than suggested by the Ki. Consequently, tiotropium derivatives devoid of the interacting hydroxy group show overproportionally short half-lives. Microsecond MD simulations show that this double hydrogen bonded interaction hinders tiotropium from moving into the exit channel by reducing the frequency of tyrosine-lid opening movements. Taken together, our data show that the interaction with N5086.52 is indeed an essential prerequisite for the development of slowly dissociating muscarinic receptor inverse agonists. This hypothesis is corroborated by the a posteriori observation that the only highly conserved substructure of all long-acting antimuscarinic agents currently in clinical development or already on the market is the hydroxy group.

But the extracellular loops also get into the act. The m2 subtype's nearby loop seems to be more flexible than the one in m3, and there's a lysine in the m3 that probably contributes some electrostatic repulsion to the charged tiotropium as it tries to back out of the protein. That's another effect that's hard to take advantage of, since the charged region of the molecule is a key for binding down in the active site, and messing with it would probably not pay dividends.

But there are some good take-aways from this paper. The authors note that the X-ray structure, while valuable, seems to have large confirmed the data generated by mutagenesis (as well it should). So if you're willing to do lots of point mutations, on both your ligand and your protein, you can (in theory) work some of these fine details out. Molecular dynamics simulations would seem to be of help here, too, also in theory. I'd be interested to hear if people can corroborate that with real-world experience.

Comments (20) + TrackBacks (0) | Category: Drug Assays | In Silico | Pharmacokinetics | The Central Nervous System

September 18, 2013

The Arguing Over PTC124 and Duchenne Muscular Dystrophy

Email This Entry

Posted by Derek

Does it matter how a drug works, if it works? PTC Therapeutics seems bent on giving everyone an answer to that question, because there sure seem to be a lot of questions about how ataluren (PTC124), their Duchenne Muscular Dystrophy (DMD) therapy, acts. This article at Nature Biotechnology does an excellent job explaining the details.

Premature "stop" codons in the DNA of DMD patients, particularly in the dystrophin gene, are widely thought to be one of the underlying problems in the disease. (The same mechanism is believed to operate in many other genetic-mutation-driven conditions as well. Ataluren is supposed to promote "read-through" of these to allow the needed protein to be produced anyway. That's not a crazy idea at all - there's been a lot of thought about ways to do that, and several aminoglycoside antibiotics have been shown to work through that mechanism. Of that class, gentamicin has been given several tries in the clinic, to ambiguous effect so far.

So screening for a better enhancer of stop codon read-through seems like it's worth a shot for a disease with so few therapeutic options. PTC did this using a firefly luciferase (Fluc) reporter assay. As with any assay, there are plenty of opportunities to get false positives and false negatives. Firefly luciferase, as a readout, suffers from instability under some conditions. And if its signal is going to wink out on its own, then a compound that stabilizes it will look like a hit in your assay system. Unfortunately, there's no particular market in humans for a compound that just stabilizes firefly luciferase.

That's where the argument is with ataluren. Papers have appeared from a team at the NIH detailing trouble with the FLuc readout. That second paper (open access) goes into great detail about the mechanism, and it's an interesting one. FLuc apparently catalyzes a reaction between PTC124 and ATP, to give a new mixed anhydride adduct that is a powerful inhibitor of the enzyme. The enzyme's normal mechanism involves a reaction between luciferin and ATP, and since luciferin actually looks like something you'd get in a discount small-molecule screening collection, you have to be alert to something like this happening. The inhibitor-FLuc complex keeps the enzyme from degrading, but the new PTC124-derived inhibitor itself is degraded by Coenzyme A - which is present in the assay mixture, too. The end result is more luciferase signal that you expect versus the controls, which looks like a hit from your reporter gene system - but isn't. PTC's scientists have replied to some of these criticisms here.

Just to add more logs to the fire, other groups have reported that PTC124 seems to be effective in restoring read-through for similar nonsense mutations in other genes entirely. But now there's another new paper, this one from a different group at Dundee, claiming that ataluren fails to work through its putative mechanism under a variety of conditions, which would seem to call these results into question as well. Gentamicin works for them, but not PTC124. Here's the new paper's take-away:

In 2007 a drug was developed called PTC124 (latterly known as Ataluren), which was reported to help the ribosome skip over the premature stop, restore production of functional protein, and thereby potentially treat these genetic diseases. In 2009, however, questions were raised about the initial discovery of this drug; PTC124 was shown to interfere with the assay used in its discovery in a way that might be mistaken for genuine activity. As doubts regarding PTC124's efficacy remain unresolved, here we conducted a thorough and systematic investigation of the proposed mechanism of action of PTC124 in a wide array of cell-based assays. We found no evidence of such translational read-through activity for PTC124, suggesting that its development may indeed have been a consequence of the choice of assay used in the drug discovery process.

Now this is a mess, and it's complicated still more by the not-so-impressive performance of PTC124 in the clinic. Here's the Nature Biotechnology article's summary:

In 2008, PTC secured an upfront payment of $100 million from Genzyme (now part of Paris-based Sanofi) in return for rights to the product outside the US and Canada. But the deal was terminated following lackluster data from a phase 2b trial in DMD. Subsequently, a phase 3 trial in cystic fibrosis also failed to reach statistical significance. Because the drug showed signs of efficacy in each indication, however, PTC pressed ahead. A phase 3 trial in DMD is now underway, and a second phase 3 trial in cystic fibrosis will commence shortly.

It should be noted that the read-through drug space has other players in it as well. Prosensa/GSK and Sarepta are in the clinic with competing antisense oligonucleotides targeting a particular exon/mutation combination, although this would probably taken them into other subpopulations of DMD patients than PTC is looking to treat.

If they were to see real efficacy, PTC could have the last laugh here. To get back to the first paragraph of this post, if a compound works, well, the big argument has just been won. The company has in vivo data to show that some gene function is being restored, as well they should (you don't advance a compound to the clinic just on the basis of in vitro assay numbers, no matter how they look). It could be that the compound is a false positive in the original assay but manages to work through some other mechanism, although no one knows what that might be.

But as you can see, opinion is very much divided about whether PTC124 works at all in the real clinical world. If it doesn't, then the various groups detailing trouble with the early assays will have a good case that this compound never should have gotten as far as it did.

Comments (28) + TrackBacks (0) | Category: Biological News | Business and Markets | Drug Assays | Drug Development

September 12, 2013

Ligands From Nothing

Email This Entry

Posted by Derek

Well, nearly nothing. That's the promise of a technique that's been published by the Ernst lab from the University of Basel. They first wrote about this in 2010, in a paper looking for ligands to the myelin-associated glycoprotein (MAG). That doesn't sound much like a traditional drug target, and so it isn't. It's part of a group of immunoglobulin-like lectins, and they bind things like sialic acids and gangliosides, and they don't seem to bind them very tightly, either.

One of these sialic acids was used as their starting point, even though its affinity is only 137 micromolar. They took this structure and hung a spin label off it, with a short chain spacer. The NMR-savvy among you will already see an application of Wolfgang Jahnke's spin-label screening idea (SLAPSTIC) coming. That's based on the effect of an unpaired electron in NMR spectra - it messes with the relaxation time of protons in the vicinity, and this can be used to determine whatever might be nearby. With the right pulse sequence, you can easily detect any protons on any other molecules or residues out to about 15 or 20 Angstroms from the spin label.

Jahnke's group at Novartis attached spin labels to proteins and used these the find ligands by NMR screening. The NMR field has a traditional bias towards bizarre acronyms, which sometimes calls for ignoring a word or two, so SLAPSTIC stands for "Spin Labels Attached to Protein Side chains as a Tool to identify Interacting Compounds". Ernst's team took their cue from yet another NMR ligand-screening idea, the Abbott "SAR by NMR" scheme. That one burst on the scene in 1996, and caused a lot of stir at the time. The idea was that you could use NMR of labeled proteins, with knowledge of their structure, to find sets of ligands at multiple binding sites, then chemically stitch these together to make a much more potent inhibitor. (This was fragment-based drug discovery before anyone was using that phrase).

The theory behind this idea is perfectly sound. It's the practice that turned out to be the hard part. While fragment linking examples have certainly appeared (including Abbott examples), the straight SAR-by-NMR technique has apparently had a very low success rate, despite (I'm told by veterans of other companies) a good deal of time, money, and effort in the late 1990s. Getting NMR-friendly proteins whose structure was worked out, finding multiple ligands at multiple sites, and (especially) getting these fragments linked together productively has not been easy at all.
But Ernst's group has brought the idea back. They did a second-site NMR screen with a library of fragments and their spin-labeled sialic thingie, and found that 5-nitroindole was bound nearby, with the 3-position pointed towards the label. That's an advantage of this idea - you get spatial and structural information without having to label the protein itself, and without having to know anything about its structure. SPR experiments showed that the nitroindole alone had affinity up in the millimolar range.

They then did something that warmed my heart. They linked the fragments by attaching a range of acetylene and azide-containing chains to the appropriate ends of the two molecules and ran a Sharpless-style in situ click reaction. I've always loved that technique, partly because it's also structure-agnostic. In this case, they did a 3x4 mixture of coupling partners, potentially forming 24 triazoles (syn and anti). After three days of incubation with the protein, a new peak showed up in the LC/MS corresponding to a particular combination. They synthesized both possible candidates, and one of them was 2 micromolar, while the other was 190 nanomolar.
That molecule is shown here - the percentages in the figure are magnetization transfer in STD experiments, with the N-acetyl set to 100% as reference. And that tells you that both ends of the molecule are indeed participating in the binding, as that greatly increased affinity would indicate. (Note that the triazole appears to be getting into the act, too). That affinity is worth thinking about - one part of this molecule was over 100 micromolar, and the other was millimolar, but the combination is 190 nanomolar. That sort of effect is why people keep coming back to fragment linking, even though it's been a brutal thing to get to work.

When I read this paper at the time, I thought that it was very nice, and I filed it in my "Break Glass in Case of Emergency" section for interesting and unusual screening techniques. One thing that worried me, as usual, was whether this was the only system this had ever worked on, or ever would. So I was quite happy to see a new paper from the Ernst group this summer, in which they did it again. This time, they found a ligand for E-selectin, another one of these things that you don't expect to ever find a decent small molecule for.

In this case, it's still not what an organic chemist would be likely to call a "decent small molecule", because they started with something akin to sialyl Lewis, which is already a funky tetrasaccharide. Their trisaccharide derivative had roughly 1 micromolar affinity, with the spin label attached. A fragment screen against E-selectrin had already identified several candidates that seemed to bind to the protein, and the best guess what that they probably wouldn't be binding in the carbohydrate recognition region. Doing the second-site screen as before gave them, as fate would have it, 5-nitroindole as the best candidate. (Now my worry is that this technique only works when you run it with 5-nitroindole. . .)

They worked out the relative geometry of binding from the NMR experiments, and set about synthesizing various azide/acetylene combinations. In this case, the in situ Sharpless-style click reactions did not give any measurable products, perhaps because the wide, flat binding site wasn't able to act as much of a catalyst to bring the two compounds together. Making a library of triazoles via the copper-catalyzed route and testing those, though, gave several compounds with affinities between 20x and 50x greater than the starting structure, and with dramatically slower off-rates.

They did try to get rid of the nitro group, recognizing that it's only an invitation to trouble. But the few modifications they tried really lowered the affinity, which tells you that the nitro itself was probably an important component of the second-site binding. That, to me, is argument enough to consider not having those things in your screening collection to start with. It all depends on what you're hoping for - if you just want a ligand to use as a biophysical tool compound, then nitro on, if you so desire. But it's hard to stop there. If it's a good hit, people will want to put it into cells, into animals, into who knows what, and then the heartache will start. If you're thinking about these kinds of assays, you might well be better off not knowing about some functionality that has a very high chance of wasting your time later on. (More on this issue here, here, here, and here). Update: here's more on trying to get rid of nitro groups).

This work, though, is the sort of thing I could read about all day. I'm very interested in ways to produce potent compounds from weak binders, ways to attack difficult low-hit-rate targets, in situ compound formation, and fragment-based methods, so these papers push several of my buttons simultaneously. And who knows, maybe I'll have a chance to do something like this all day at some point. It looks like work well worth taking seriously.

Comments (20) + TrackBacks (0) | Category: Analytical Chemistry | Chemical News | Drug Assays

September 4, 2013

More Thoughts on Compound Metrics

Email This Entry

Posted by Derek

Over at Practical Fragments, Dan Erlanson has comments on the Michael Shultz paper that I wrote about here. He goes into details on some of the problems that turn up when you try to apply various compound metrics across a broad range of molecular weights and/or lipophilicities. In the most obvious example, the indices that are based on Heavy Atom Count (HAC) will jump around much more in the low-molecular-weight range, and none of the proposed refinements can quite fix this. And with the alternative LELP measure, you have to watch out when you're at very low LogP values.

Shultz's preferred LipE/LLE metric avoids that problem, and it's size-independent as well. That part can be either a bug or a feature, depending on your perspective. For the most part, I think that's useful, but in the early stages of fragment optimization, I think that a size-independent measurement is not what you want. The whole point in that stage is to pick the starting points with the most binding for their size, and a well-designed fragment library shouldn't have too many big problems with lipophilicity (those will come along later). So I take Shultz's point about the validity of LLE in general, but I think that I'll be using it and either LE or BEI (HAC-driven and molecular-weight driven) measure of binding efficiency when I'm working in the fragment end of things. How to weight those will be a judgment call, naturally, but judgment calls are, in theory, what we're being paid for, right?

Comments (3) + TrackBacks (0) | Category: Drug Assays | In Silico

August 23, 2013

Chemistry On The End of DNA

Email This Entry

Posted by Derek

We chemists have always looked at the chemical machinery of living systems with a sense of awe. A billion years of ruthless pruning (work, or die) have left us with some bizarrely efficient molecular catalysts, the enzymes that casually make and break bonds with a grace and elegance that our own techniques have trouble even approaching. The systems around DNA replication are particularly interesting, since that's one of the parts you'd expect to be under the most selection pressure (every time a cell divides, things had better work).

But we're not content with just standing around envying the polymerase chain reaction and all the rest of the machinery. Over the years, we've tried to borrow whatever we can for our own purposes - these tools are so powerful that we can't resist finding ways to do organic chemistry with them. I've got a particular weakness for these sorts of ideas myself, and I keep a large folder of papers (electronic, these days) on the subject.

So I was interested to have a reader send along this work, which I'd missed when it came out on PLOSONE. It's from Pehr Harbury's group at Stanford, and it's in the DNA-linked-small-molecule category (which I've written about, in other cases, here and here). Here's a good look at the pluses and minuses of this idea:

However, with increasing library complexity, the task of identifying useful ligands (the ‘‘needles in the haystack’’) has become increasingly difficult. In favorable cases, a bulk selection for binding to a target can enrich a ligand from non-ligands by about 1000-fold. Given a starting library of 1010 to 1015 different compounds, an enriched ligand will be present at only 1 part in 107 to 1 part in 1012. Confidently detecting such rare molecules is hard, even with the application of next-generation sequencing techniques. The problem is exacerbated when biologically-relevant selections with fold-enrichments much smaller than 1000-fold are utilized.

Ideally, it would be possible to evolve small-molecule ligands out of DNA-linked chemical libraries in exactly the same way that biopolymer ligands are evolved from nucleic acid and protein libraries. In vitro evolution techniques overcome the ‘‘needle in the haystack’’ problem because they utilize multiple rounds of selection, reproductive amplification and library re-synthesis. Repetition provides unbounded fold-enrichments, even for inherently noisy selections. However, repetition also requires populations that can self-replicate.

That it does, and that's really the Holy Grail of evolution-linked organic synthesis - being able to harness the whole process. In this sort of system, we're talking about using the DNA itself as a physical prod for chemical reactivity. That's also been a hot field, and I've written about some examples from the Liu lab at Harvard here, here, and here. But in this case, the DNA chemistry is being done with all the other enzymatic machinery in place:

The DNA brings an incipient small molecule and suitable chemical building blocks into physical proximity and induces covalent bond formation between them. In so doing, the naked DNA functions as a gene: it orchestrates the assembly of a corresponding small molecule gene product. DNA genes that program highly fit small molecules can be enriched by selection, replicated by PCR, and then re-translated into DNA-linked chemical progeny. Whereas the Lerner-Brenner style DNA-linked small-molecule libraries are sterile and can only be subjected to selective pressure over one generation, DNA-programmed libraries produce many generations of offspring suitable for breeding.

The scheme below shows how this looks. You take a wide variety of DNA sequences, and have them each attached to some small-molecule handle (like a primary amine). You then partition these out into groups by using resins that are derivatized with oligonucleotide sequences, and you plate these out into 384-well format. While the DNA end is stuck to the resin, you do chemistry on the amine end (and the resin attachment lets you get away with stuff that would normally not work if the whole DNA-attached thing had to be in solution). You put a different reacting partner in each of the 384 wells, just like in the good ol' combichem split/pool days, just with DNA as the physical separation mechanism.
In this case, the group used 240-base-pair DNA sequences, two hundred seventeen billion of them. That sentence is where you really step off the edge into molecular biology, because without its tools, generating that many different species, efficiently and in usable form, is pretty much out of the question with current technology. That's five different coding sequences, in their scheme, with 384 different ones in each of the first four (designated A through D), and ten in the last one, E. How diverse was this, really? Get ready for more molecular biology tools:

We determined the sequence of 4.6 million distinct genes from the assembled library to characterize how well it covered ‘‘genetic space’’. Ninety-seven percent of the gene sequences occurred only once (the mean sequence count was 1.03), and the most abundant gene sequence occurred one hundred times. Every possible codon was observed at each coding position. Codon usage, however, deviated significantly from an expectation of random sampling with equal probability. The codon usage histograms followed a log-normal distribution, with one standard deviation in log- likelihood corresponding to two-to-three fold differences in codon frequency. Importantly, no correlation existed between codon identities at any pair of coding positions. Thus, the likelihood of any particular gene sequence can be well approxi- mated by the product of the likelihoods of its constituent codons. Based on this approximation, 36% of all possible genes would be present at 100 copies or more in a 10 picomole aliquot of library material, 78% of the genes would be present at 10 copies or more, and 4% of the genes would be absent. A typical selection experiment (10 picomoles of starting material) would thus sample most of the attainable diversity.

The group had done something similar before with 80-codon DNA sequences, but this system has 1546, which is a different beast. But it seems to work pretty well. Control experiments showed that the hybridization specificity remained high, and that the micro/meso fluidic platform being used could return products with high yield. A test run also gave them confidence in the system: they set up a run with all the codons except one specific dropout (C37), and also prepared a "short gene", containing the C37 codon, but lacking the whole D area (200 base pairs instead of 240). When they mixed that in with the drop-out library (in a ratio of 1 to 384), and split that out onto a C-codon-attaching array of beads. They then did the chemical step, attaching one peptoid piece onto all of them except the C37 binding well - that one got biotin hydrazide instead. Running the lot of them past streptavidin took the ratio of the C37-containing ones from 1:384 to something over 35:1, an enhancement of at least 13,000-fold. (Subcloning and sequencing of 20 isolates showed they all had the C37 short gene in them, as you'd expect).

They then set up a three-step coupling of peptoid building blocks on a specific codon sequence, and this returned very good yields and specificities. (They used a fluorescein-tagged gene and digested the product with PDE1 before analyzing them at each step, which ate the DNA tags off of them to facilitate detection). The door, then, would now seem to be open:

Exploration of large chemical spaces for molecules with novel and desired activities will continue to be a useful approach in academic studies and pharmaceutical investigations. Towards this end, DNA-programmed combinatorial chemistry facilitates a more rapid and efficient search process over a larger chemical space than does conventional high-throughput screening. However, for DNA-programmed combinatorial chemistry to be widely adopted, a high-fidelity, robust and general translation system must be available. This paper demonstrates a solution to that challenge.

The parallel chemical translation process described above is flexible. The devices and procedures are modular and can be used to divide a degenerate DNA population into a number of distinct sub-pools ranging from 1 to 384 at each step. This coding capacity opens the door for a wealth of chemical options and for the inclusion of diversity elements with widely varying size, hydrophobicity, charge, rigidity, aromaticity, and heteroatom content, allowing the search for ligands in a ‘‘hypothesis-free’’ fashion. Alternatively, the capacity can be used to elaborate a variety of subtle changes to a known compound and exhaustively probe structure-activity relationships. In this case, some elements in a synthetic scheme can be diversified while others are conserved (for example, chemical elements known to have a particular structural or electrostatic constraint, modular chemical fragments that independently bind to a protein target, metal chelating functional groups, fluorophores). By facilitating the synthesis and testing of varied chemical collections, the tools and methods reported here should accelerate the application of ‘‘designer’’ small molecules to problems in basic science, industrial chemistry and medicine.

Anyone want to step through? If GSK is getting some of their DNA-coded screening to work (or at least telling us about the examples that did?), could this be a useful platform as well? Thoughts welcome in the comments.

Comments (13) + TrackBacks (0) | Category: Chemical Biology | Chemical News | Drug Assays

August 22, 2013

Too Many Metrics

Email This Entry

Posted by Derek

Here's a new paper from Michael Shultz of Novartis, who is trying to cut through the mass of metrics for new compounds. I cannot resist quoting his opening paragraph, but I do not have a spare two hours to add all the links:

Approximately 15 years ago Lipinski et al. published their seminal work linking molecular properties with oral absorption.1 Since this ‘Big Bang’ of physical property analysis, the universe of parameters, rules and optimization metrics has been expanding at an ever increasing rate (Figure 1).2 Relationships with molecular weight (MW), lipophilicity,3 and 4 ionization state,5 pKa, molecular volume and total polar surface area have been examined.6 Aromatic rings,7 and 8 oxygen atoms, nitrogen atoms, sp3 carbon atoms,9 chiral atoms,9 non-hydrogen atoms, aromatic versus non-hydrogen atoms,10 aromatic atoms minus sp3 carbon atoms,6 and 11 hydrogen bond donors, hydrogen bond acceptors and rotatable bonds12 have been counted and correlated.13 In addition to the rules of five came the rules of 4/40014 and 3/75.15 Medicinal chemists can choose from composite parameters (or efficiency indices) such as ligand efficiency (LE),16 group efficiency (GE), lipophilic efficiency/lipophilic ligand efficiency (LipE17/LLE),18 ligand lipophilicity index (LLEAT),19 ligand efficiency dependent lipophilicity (LELP), fit quality scaled ligand efficiency (LE_scale),20 percentage efficiency index (PEI),21 size independent ligand efficiency (SILE), binding efficiency index (BEI) or surface binding efficiency index (SEI)22 and composite parameters are even now being used in combination.23 Efficiency of binding kinetics has recently been introduced.24 A new trend of anthropomorphizing molecular optimization has occurred as molecular ‘addictions’ and ‘obesity’ have been identified.25 To help medicinal chemists there are guideposts,21 rules of thumb,14 and 26 a property forecast index,27 graphical representations of properties28 such as efficiency maps, atlases,29 ChemGPS,30 traffic lights,31 radar plots,32 Craig plots,33 flower plots,34 egg plots,35 time series plots,36 oral bioavailability graphs,37 face diagrams,28 spider diagrams,38 the golden triangle39 and the golden ratio.40

He must have enjoyed writing that one, if not tracking down all the references. This paper is valuable right from the start just for having gathered all this into one place! But as you read on, you find that he's not too happy with many of these metrics - and since there's no way that they can all be equally correct, or equally useful, he sets himself the task of figuring out which ones we can discard. The last reference in the quoted section below is to the famous "Can a biologist fix a radio?" paper:

While individual composite parameters have been developed to address specific relationships between properties and structural features (e.g. solubility and aromatic ring count) the benefit may be outweighed by the contradictions that arise from utilizing several indices at once or the complexity of adopting and abandoning various metrics depending on the stage of molecular optimization. The average medicinal chemist can be overwhelmed by the ‘analysis fatigue’ that this plethora of new and contradictory tools, rules and visualizations now provide, especially when combined with the increasing number of safety, off-target, physicochemical property and ADME data acquired during optimization efforts. Decision making is impeded when evaluating information that is wrong or excessive and thus should be limited to the absolute minimum and most relevant available.

As Lazebnik described, sometimes the more facts we learn, the less we understand.

And he discards quite a few. All the equations that involve taking the log of potency and dividing by the heavy atom count (HAC), etc., are playing rather loose with the math:

To be valid, LE must remain constant for each heavy atom that changes potency 10-fold. This is not the case as a 15 HAC compound with a pIC50 of 3 does not have the same LE as a 16 HAC compound with a pIC50 of 4 (ΔpIC50 = 1, ΔHAC = 1, ΔLE = 0.07). A 10-fold change in potency per heavy atom does not result in constant LE as defined by Hopkins, nor will it result in a constant SILE, FQ or LLEAT values. These metrics do not mathematically normalize size or potency because they violate the quotient rule of logarithms. To obey this rule and be a valid mathematical function HAC would subtracted from pIC50 and rendered independent of size and reference potency.

Note that he's not recommending that last operation as a guideline, either. Another conceptual problem with plain heavy atom counting is that it treats all atoms the same, but that's clearly an oversimplification. But dividing by some form of molecular weight is an oversimplification, too: a nitrogen differs from an oxygen by a lot more than that 1 mass unit. (This topic came up here a little while back). But oversimplified or not - heck, mathematically valid or not - the question is whether these things help out enough when used as metrics in the real world. And Shultz would argue that they don't. Keeping LE the same (or even raising it) is supposed to be the sign of a successful optimization, but in practice, LE usually degrades. His take on this is that "Since lower ligand efficiency is indicative of both higher and lower probabilities of success (two mutually exclusive states) LE can be invalidated by not correlating with successful optimization."

I think that's too much of a leap - because successful drug programs have had their LE go down during the process, that doesn't mean that this was a necessary condition, or that they should have been aiming for that. Perhaps things would have been even better if they hadn't gone down (although I realize that arguing from things that didn't happen doesn't have much logical force). Try looking at it this way: a large number of successful drug programs have had someone high up in management trying to kill them along the way, as have (obviously) most of the unsuccessful ones. That would mean that upper management decisions to kill a program are also indicative of both higher and lower probabilities of success, and can thus be invalidated, too. Actually, he might be on to something there.

Shultz, though, finds that he's not able to invalidate LipE (or LLE), variously known as ligand-lipophilicity efficiency or lipophilic ligand efficiency. That's p(IC50) - logP, which at least follows the way that logarithms of quotients are supposed to work. And it also has been shown to improve during known drug optimization campaigns. The paper has a thought experiment, on some hypothetical compounds, as well as some data from a tankyrase inhibitor series that seem to show the LipE behave more rationally than other metrics (which sometimes start pointing in opposite directions).

I found the chart below to be quite interesting. It uses the cLogP data from Paul Leeson and Brian Springthorpe's original LLE paper (linked in the above paragraph) to show what change in potency you would expect when you change a hydrogen in your molecule to one of the groups shown if you're going to maintain a constant LipE value. So while hydrophobic groups tend to make things more potent, this puts a number on it. A t-butyl, for example, should make things about 50-fold more potent if it's going to pull its weight as a ball of grease. (Note that we're not talking about effects on PK and tox here, just sheer potency - if you play this game, though, you'd better be prepared to keep an eye on things downstream).
On the other end of the scale, a methoxy should, in theory, cut your potency roughly in half. If it doesn't, that's a good sign. A morpholine should be three or four times worse, and if it isn't, then it's found something at least marginally useful to do in your compound's binding site. What we're measuring here is the partitioning between your compound wanting to be in solution, and wanting to be in the binding site. More specifically, since logP is in the equation, we're looking at the difference in the partitioning of your compound between octanol and water, versus its partitioning between the target protein and water. I think we can all agree that we'd rather have compounds that bind because they like something about the active site, rather than just fleeing the solution phase.

So in light of this paper, I'm rethinking my ligand-efficiency metrics. I'm still grappling with how LipE performs down at the fragment end of the molecular weight scale, and would be glad to hear thoughts on that. But Shultz's paper, if it can get us to toss out a lot of the proposed metrics already in the literature, will have done us all a service.

Comments (38) + TrackBacks (0) | Category: Drug Assays | Drug Development | In Silico | Pharmacokinetics

August 20, 2013

GPCRs As Drug Targets: Nowhere Near Played Out

Email This Entry

Posted by Derek

Here's a paper that asks whether GPCRs are still a source of new targets. As you might guess, the answer is "Yes, indeed". (Here's a background post on this area from a few years ago, and here's my most recent look at the area).

It's been a famously productive field, but the distribution is pretty skewed:

From a total of 1479 underlying targets for the action of 1663 drugs, 109 (7%) were GPCRs or GPCR related (e.g., receptor-activity modifying proteins or RAMPs). This immediately reveals an issue: 26% of drugs target GPCRs, but they account for only 7% of the underlying targets. The results are heavily skewed by certain receptors that have far more than their “fair share” of drugs. The most commonly targeted receptors are as follows: histamine H1 (77 occurrences), α1A adrenergic (73), muscarinic M1 (72), dopamine D2 (62), muscarinic M2 (60), 5HT2a (59), α2A adrenergic (56), and muscarinic M3 (55)—notably, these are all aminergic GPCRs. Even the calculation that the available drugs exert their effects via 109 GPCR or GPCR-related targets is almost certainly an overestimate since it includes a fair proportion where there are only a very small number of active agents, and they all have a pharmacological action that is “unknown”; in truth, we have probably yet to discover an agent with a compelling activity at the target in question, let alone one with exactly the right pharmacology and appropriately tuned pharmacokinetics (PK), pharmacodynamics (PD), and selectivity to give clinical efficacy for our disease of choice. A prime example of this would be the eight metabotropic (mGluR) receptors, many of which have only been “drugged” according to this analysis due to the availability of the endogenous ligand (L-glutamic acid) as an approved nutraceutical. There are also a considerable number of targets for which the only known agents are peptides, rather than small molecules. . .

Of course, since we're dealing with cell-surface receptors, peptides (and full-sized proteins) have a better shot at becoming drugs in this space.

Of the 437 drugs found to target GPCRs, 21 are classified as “biotech” (i.e., biopharmaceuticals) with the rest as “small molecules.” However, that definition seems rather generous given that the molecular weight (MW) of the “small molecules” extends as high as 1623. Using a fairly modest threshold of MW <600 suggests that ~387 are more truly small molecules and ~50 are non–small molecules, being roughly an 80:20 split. Pursuing the 20%, while not being novel targets/mechanisms, could still provide important new oral/small-molecule medications with the comfort of excellent existing clinical validation. . .

The paper goes on to mention many other possible modes for drug action - allosteric modulators, GPCR homo- and heterodimerization, other GPCR-protein interactions, inverse agonists and the like, alternative signaling pathways other than the canonical G-proteins, and more. It's safe to say that all this will keep up busy for a long time to come, although working up reliable assays for some of these things is no small matter.

Comments (4) + TrackBacks (0) | Category: Biological News | Drug Assays

August 19, 2013

High Throughput Screening Services

Email This Entry

Posted by Derek

Here's a question that comes up once in a while in my e-mail. I've always worked for companies that are large enough to do all of their own high-throughput screening (with some exceptions for when we've tried out some technology that we don't have in-house, or not yet). But there are many smaller companies that contract out for some or all of their screening, and sometimes for some assay development as well beforehand.

So there are, naturally, plenty of third parties who will run screens for you, against their own compound collections or against something you bring them. A reader was just asking me if I had any favorites in this area myself, but I haven't had enough call to use these folks to have a useful opinion. So I think it would be worth hearing about experiences with these shops, good and bad. Keep the specific recommendations recent, if possible, but general advice and What Not to Do warnings are timeless. Any thoughts?

Comments (33) + TrackBacks (0) | Category: Drug Assays

August 8, 2013

The 3D Fragment Consortium

Email This Entry

Posted by Derek

Fragment-based screening comes up here fairly often (and if you're interested in the field, you should also have Practical Fragments on your reading list). One of the complaints both inside and outside the fragment world is that there are a lot of primary hits that fall into flat/aromatic chemical space (I know that those two don't overlap perfectly, but you know the sort of things I mean). The early fragment libraries were heavy in that sort of chemical matter, and the sort of collections you can buy still tend to be.

So people have talked about bringing in natural-product-like structures, and diversity-oriented-synthesis structures and other chemistries that make more three-dimensional systems. The commercial suppliers have been catching up with this trend, too, although some definitions of "three-dimensional" may not match yours. (Does a biphenyl derivative count, or is that what you're trying to get away from?)

The UK-based 3D Fragment Consortium has a paper out now in Drug Discovery Today that brings together a lot of references to work in this field. Even if you don't do fragment-based work, I think you'll find it interesting, because many of the same issues apply to larger molecules as well. How much return do you get for putting chiral centers into your molecules, on average? What about molecules with lots of saturated atoms that are still rather squashed and shapeless, versus ones full of aromatic carbons that carve out 3D space surprisingly well? Do different collections of these various molecular types really have differences in screening hit rates, and do these vary by the target class you're screening against? How much are properties (solubility, in particular) shifting these numbers around? And so on.

The consortium's site is worth checking out as well for more on their activities. One interesting bit of information is that the teams ended up crossing off over 90% of the commercially available fragments due to flat structures, which sounds about right. And that takes them where you'd expect it to:

We have concluded that bespoke synthesis, rather than expansion through acquisition of currently available commercial fragment-sized compounds is the most appropriate way to develop the library to attain the desired profile. . .The need to synthesise novel molecules that expand biologically relevant chemical space demonstrates the significant role that academic synthetic chemistry can have in facilitating target evaluation and generating the most appropriate start points for drug discovery programs. Several groups are devising new and innovative methodologies (i.e. methyl activation, cascade reactions and enzymatic functionalisation) and techniques (e.g. flow and photochemistry) that can be harnessed to facilitate expansion of drug discovery-relevant chemical space.

And as long as they stay away from the frequent hitters/PAINS, they should end up with a good collection. I look forward to future publications from the group to see how things work out!

Comments (3) + TrackBacks (0) | Category: Analytical Chemistry | Chemical News | Drug Assays | In Silico

July 29, 2013

More Whitesides on Ligand Binding

Email This Entry

Posted by Derek

George Whitesides and his lab have another paper out on the details of how ligands bind to proteins. They're still using the favorite model enzyme of all time (carbonic anhydrase), the fruit fly and nematode of the protein world. Last time around, using a series of ligands and their analogs with an extra phenyl in their structure. The benzo-ligands had increased affinity, and this seemed to be mostly an enthalpic effect. After a good deal of calorimetry, etc., they concluded that the balancing act between enthalpy and entropy they saw over the group was different for ligand binding than it was for logP partitioning, and that means that it doesn't really match up with the accepted definition of a "hydrophobic effect".

In this study, they're looking at fluorinated analogs of the same compounds to see what that might do to the binding process. That makes the whole thing interesting for a medicinal chemist, because we make an awful lot of fluorinated analogs. You can start some interesting discussions about whether these are more hydrophobic than their non-F analogs, though, and this paper lands right in the middle of that issue.

The first result was that the fluorinated analogs bound to the enzyme (in their X-ray structures) with almost identical geometry. That makes the rest of the discussion easier to draw conclusions from (and more relevant). It's worth remembering, though, that very small changes can still add up. There was a bit of a shift in the binding pocket, actually, which they attribute to an unfavorable interaction between the fluorines and the carbonyl of a threonine residue. But the carbonic anhydrase pocket is pretty accomodating - the overall affinity of the compounds did not really change. That led to this conclusion:

Values of DG8bind, combined with an overall conserved binding geometry of each set of benzo- and fluorobenzo-extended ligands suggest that binding depends on a fine balance of interactions between HCA, the ligand, and molecules of water filling the pocket and surrounding the ligand, and that a simple analysis of interactions between the protein and ligand (Figure1E) is insufficient to understand (or more importantly, predict) the free energy of binding.

But although the overall free energy didn't change, the enthalpic and entropic components did (but arrived at the same place, another example to add to the long list of systems that do this). The differences seem to be in the Coulombic interaction with the binding pocket (worse enthalpy term - is that what shifted the structure over a bit in the X-ray?) and changes in energy of solvation as the ligand binds (better entropy term). Matched pairs of compounds didn't really show a difference in how many waters they displaced from the binding site.

So the take-home is that the hydrophobic effect is not all about releasing waters from protein binding surfaces, as has been proposed by some. It's a mixture of stuff, and especially depends on the structure of the water in the binding pocket and around the ligands, and the changes in these as the compounds leave bulk solvent and find their way into the binding site.

That makes things tricky for many compounds. Hydrophobic effects seem to be a big part of the binding energy of a lot of drug molecules (despite various efforts to cut back on this), and these Whitesides studies would say that modeling and predicting these energetic changes are going to be hard. Computationally, we'd have an easier time figuring out direct interactions between the protein and the ligand, the way we do with enthalpic interactions like hydrogen bonds. Keeping track of all those water molecules is more painful - but necessary.

Comments (10) + TrackBacks (0) | Category: Drug Assays | In Silico

June 26, 2013

Protein-Protein Compounds - The Flip Side

Email This Entry

Posted by Derek

The topic of protein-protein inhibitor compounds has come up around here several times. It's the classic "undruggable" target, although that adjective isn't quite accurate. Let's leave it at "definitely harder than the usual stuff"; no one could argue with that.

But there's a flip side to this area that people don't think about so much. What about a compound that would make two proteins interact more tightly? A conversation with a reader of the site got me to thinking about this, and it turns out that there's a good review of the concept here, from 2012. The compounds that are known to really do this sort of thing all seem to be natural products, which I don't suppose should come as a surprise. The most well-worked-out of the group is (as some readers will have guessed) FK506 (tacrolimus). Very few drug research organizations have been brave enough to tackle a mechanism like this, so you're not going to see many examples of synthetic compounds. How small (and drug-like) a compound can be and still work through a mechanism like this is an open question.

In principle, it shouldn't be that hard a screen to run - you could imagine an assay where you watch a FRET signal hang around instead of disappearing (once you're sure that hanging all the FRET thingies off the protein partners didn't mess with the binding event, of course). You'd probably be able to see this effect by biophysical techniques as well - NMR, SPR (if you could recapitulate the protein-protein interaction with an immobilized partner on a chip), etc. You'd want a lot of structural information - seeing some sort of plausible binding surface that spans the two proteins would help to settle the nerves a bit.

You'd also want some targets, but there are probably more of them than we're used to thinking about. That's because we're don't tend to think about this mode of action at all, and if you're not keeping it in mind, you won't spot opportunities for it. The whole gain-of-function side of the business is hard to work in, for good reasons. I'm not aware of endogenous small molecules that work this way, so it's not like there are a lot of highly evolved binding pockets waiting for us to fill them. Come to think of it, I'm not aware of endogenous small molecules that work as protein-protein inhibitors, either - those processes seem to get regulated by modifications on the proteins themselves, by local concentration, or by intervention of still other proteins to rearrange binding surfaces. The scarce evolutionary record of this sort of thing might be an accident, or it might be telling us (believably) that this isn't an easy thing to do.

So I would not necessarily pin all my hopes for next year's new targets portfolio on one of these, but it would be interesting to screen and see what might turn up. Who wants to be first?

Update: here's an example from the recent literature for you!

Comments (27) + TrackBacks (0) | Category: Drug Assays | Natural Products

June 17, 2013

Aggravating Aggregators

Email This Entry

Posted by Derek

Compound aggregation is a well-known problem in biochemical assays (although if you go back a few years, that certainly wasn't the case). Some small molecules will start to bunch up under some assay conditions, and instead of your target protein getting inhibited by a single molecule of your test compound, the protein could look as if it's been inhibited by virtue of being dragged into a huge sticky clump of Test Compound Aggregate.

A group at Boehringer Ingleheim has a paper out in J. Med. Chem. suggesting a simple NMR readout to see if a given compound is showing aggregation behavior. It looks useful, but there's one thing I would add to it. The authors mention that they used a simple sodium phosphate buffer for their experiments, and that similar trends were observed in others (for a "limited set of compounds"). But I've heard Tony Giannetti of Genentech speak on this subject before (with reference to his specialty, surface plasmon resonance assays), and he's been pretty adamant about how situation-dependent aggregation can be.

The Shoichet lab's "Aggregator Advisor" page agrees. My worry is that some people might read this new paper and be tempted to clean their screening sets out up front, but you could throw some useful compounds out that way. But aggregation, annoyingly, appears to be a case-by-case thing. Probably the best ways to guard against it are (1) see if your assay can be run with detergent in it to start with, and be prepared to vary the amount, and (2) take your screening hits of interest and check them out individually before you decide that you're on to something. This new NMR assay would be a good way to do that, using the buffer that your screen was run in.

Another note that comes up in all discussions of aggregators is that while many of them are condition-specific, others have a wider range. Many "frequent hitter" compounds turn out to aggregate under a variety of conditions. In that case (because you've got empirical data from your own assays), it's really worth going back and flagging those things. It would seem worthwhile to go through any screening collection and pitch out the individual compounds that show up time and time again, since these are surely less likely to lead to anything useful. Some of these will, on closer inspection, turn out to be promiscuous aggregators, but there are other mechanisms for nastiness as well. In extreme cases, whole structural motifs should be given the fishy eye.

Comments (6) + TrackBacks (0) | Category: Drug Assays

June 7, 2013

Making Peroxides, Quietly And Unhelpfully

Email This Entry

Posted by Derek

Here's a problem with screening collections that I have to admit I wasn't aware of: generation of hydrogen peroxide. This paper (free access) gives an excellent overview of what's going on. Turns out that some compounds can undergo redox-cycling in the presence of the common buffer additive DTT (dithiothreitol - note - fixed brain spasm on earlier name), spitting out, in the end, a constant trickle of peroxide.

Now, for many assays, this might not mean much one way or another. But enzymes with a crucial cysteine residue are another matter. Those can get oxidized, which is irritating in these cases, because DTT is added to such assays just to keep that sort of thing from happening. That link above describes a useful horseradish peroxidase/phenol red assay to detect hydrogen peroxide generation, and its use to profile the NIH's Small Molecule Repository compound collections.

Fortunately, only a limited number of compounds have the ability to hose up your assays in this manner. Of the roughly 196,000 compounds screened, only 37 were true peroxide-generators. Quinones are serial offenders, as any chemist might expect, but if you let you screening collection fill up with quinones you have only yourself to blame. There are less obvious candidates, though: several arylsulfonamides also showed this behavior, and while those aren't everyone's favorite compounds, I'd like to see the large screening set that doesn't have some in there somewhere. It's worth noting, though, that many of the sulfonamides that were identified are also quinon-ish.

So I think the take-home advice here is to be aware if your target is sensitive to this sort of thing. Cysteine proteases are obvious candidates, but Trp can be oxidized, too, and a lot of proteins have crucial disulfides that might get unraveled. Once you've flagged your protein as a concern, be sure to run the hits you get back through this peroxide assay to make sure that you're not being led on. Trying to eliminate compounds by structural class up front is another approach, but the compounds that are first on the list are compounds that you should have trashcanned already.

Comments (7) + TrackBacks (0) | Category: Drug Assays

May 15, 2013

GSK's Published Kinase Inhibitor Set

Email This Entry

Posted by Derek

Speaking about open-source drug discovery (such as it is) and sharing of data sets (such as they are), I really should mention a significant example in this area: the GSK Published Kinase Inhibitor Set. (It was mentioned in the comments to this post). The company has made 367 compounds available to any academic investigator working in the kinase field, as long as they make their results publicly available (at ChEMBL, for example). The people at GSK doing this are David Drewry and William Zuercher, for the record - here's a recent paper from them and their co-workers on the compound set and its behavior in reporter-gene assays.

Why are they doing this? To seed discovery in the field. There's an awful lot of chemical biology to be done in the kinase field, far more than any one organization could take on, and the more sets of eyes (and cerebral cortices) that are on these problems, the better. So far, there have been about 80 collaborations, mostly in Europe and North America, all the way from broad high-content phenotypic screening to targeted efforts against rare tumor types.

The plan is to continue to firm up the collection, making more data available for each compound as work is done on them, and to add more compounds with different selectivity profiles and chemotypes. Now, the compounds so far are all things that have been published on by GSK in the past, obviating concerns about IP. There are, though, a multitude of other compounds in the literature from other companies, and you have to think that some of these would be useful additions to the set. How, though, does one get this to happen? That's the stage that things are in now. Beyond that, there's the possibility of some sort of open network to optimize entirely new probes and tools, but there's plenty that could be done even before getting to that stage.

So if you're in academia, and interested in kinase pathways, you absolutely need to take a look at this compound set. And for those of us in industry, we need to think about the benefits that we could get by helping to expand it, or by starting similar efforts of our own in other fields. The science is big enough for it. Any takers?

Comments (22) + TrackBacks (0) | Category: Academia (vs. Industry) | Biological News | Chemical News | Drug Assays

May 10, 2013

Why Not Share More Bioactivity Data?

Email This Entry

Posted by Derek

The ChEMBL database of compounds has been including bioactivity data for some time, and the next version of it is slated to have even more. There are a lot of numbers out in the open literature that can be collected, and a lot of numbers inside academic labs. But if you want to tap the deepest sources of small-molecule biological activity data, you have to look to the drug industry. We generate vast heaps of such; it's the driveshaft of the whole discovery effort.

But sharing such data is a very sticky issue. No one's going to talk about their active projects, of course, but companies are reluctant to open the books even to long-dead efforts. The upside is seen as small, and the downside (though unlikely) is seen as potentially large. Here's a post from the ChEMBL blog that talks about the problem:

. . .So, what would your answer be if someone asked you if you consider it to be a good idea if they would deposit some of their unpublished bioactivity data in ChEMBL? My guess is that you would be all in favour of this idea. 'Go for it', you might even say. On the other hand, if the same person would ask you what you think of the idea to deposit some of ‘your bioactivity data’ in ChEMBL the situation might be completely different.

First and foremost you might respond that there is no such bioactivity data that you could share. Well let’s see about that later. What other barriers are there? If we cut to the chase then there is one consideration that (at least in my experience) comes up regularly and this is the question: 'What’s in it for me?' Did you ask yourself the same question? If you did and you were thinking about ‘instant gratification’ I haven’t got a lot to offer. Sorry, to disappoint you. However, since when is science about ‘instant gratification’? If we would all start to share the bioactivity data that we can share (and yes, there is data that we can share but don’t) instead of keeping it locked up in our databases or spreadsheets this would make a huge difference to all of us. So far the main and almost exclusive way of sharing bioactivity data is through publications but this is (at least in my view) far too limited. In order to start to change this (at least a little bit) the concept of ChEMBL supplementary bioactivity data has been introduced (as part of the efforts of the Open PHACTS project,

There's more on this in an article in Future Medicinal Chemistry. Basically, if an assay has been described in an open scientific publication, the data generated through it qualifies for deposit in ChEMBL. No one's asking for companies to throw open their books, but even when details of a finished (or abandoned) project are published, there are often many more data points generated than ever get included in the manuscript. Why not give them a home?

I get the impression, though, that GSK is the only organization so far that's been willing to give this a try. So I wanted to give it some publicity as well, since there are surely many people who aren't aware of the effort at all, and might be willing to help out. I don't expect that data sharing on this level is going to lead to any immediate breakthroughs, of course, but even though assay numbers like this have a small chance of helping someone, they have a zero chance of helping if they're stuck in the digital equivalent of someone's desk drawer.

What can be shared, should be. And there's surely a lot more that falls into that category than we're used to thinking.

Comments (18) + TrackBacks (0) | Category: Drug Assays | The Scientific Literature

May 3, 2013

Drug Assay Numbers, All Over the Place

Email This Entry

Posted by Derek

There's a truly disturbing paper out in PLoSONE with potential implications for a lot of assay data out there in the literature. The authors are looking at the results of biochemical assays as a function of how the compounds are dispensed in them, pipet tip versus acoustic, which is the sort of idea that some people might roll their eyes at. But people who've actually done a lot of biological assays may well feel a chill at the thought, because this is just the sort of you're-kidding variable that can make a big difference.

Dispensing and dilution processes may profoundly influence estimates of biological activity of compounds. Published data show Ephrin type-B receptor 4 IC50 values obtained via tip-based serial dilution and dispensing versus acoustic dispensing with direct dilution differ by orders of magnitude with no correlation or ranking of datasets.

Lovely. There have been some alarm bells sounded before about disposable-pipet-tip systems. The sticky-compound problem is always out there, where various substances decide that they like the plastic walls of the apparatus a lot more than they like being in solution. That'll throw your numbers all over the place. And there have been concerns about bioactive substances leaching out of the plastic. (Those are just two recent examples - this new paper has several other references, if you're worried about this sort of thing).

This paper seems to have been set off by two recent AstraZeneca patents on the aforementioned EphB4 inhibitors. In the assay data tables, these list assay numbers as determined via both dispensing techniques, and they are indeed all over the place. One of the authors of this new paper is from Labcyte, the makers of the acoustic dispensing apparatus, and it's reasonable to suppose that their interactions with AZ called their attention to this situation. It's also reasonable to note that Labcyte itself has an interest in promoting acoustic dispensing technology, but that doesn't make the numbers any different. The fourteen compounds shown are invariably less potent via the classic pipet method, but by widely varying factors. So, which numbers are right?

The assumption would be that the more potent values have a better chance of being correct, because it's a lot easier to imagine something messing up the assay system than something making it read out at greater potency. But false positives certainly exist, too, so the authors used the data set to generate a possible pharmacophore for the compound series using both sets of numbers. And it turns out that the one from the acoustic dispensing runs gives you a binding model that matches pretty well with reality, while if you use the pipet data you get something broadly similar, but missing some important contributions from hydrophobic groups. That, plus the fact that the assay data shows a correlation with logP in the acoustic-derived data (but not so much with the pipet-derived numbers) makes it look like the sticky-compound effect might be what's operating here. But it's hard to be sure:

No previous publication has analyzed or compared such data (based on tip-based and acoustic dispensing) using computational or statistical approaches. This analysis is only possible in this study because there is data for both dispensing approaches for the compounds in the patents from AstraZeneca that includes molecule structures. We have taken advantage of this small but valuable dataset to perform the analyses described. Unfortunately it is unlikely that a major pharmaceutical company will release 100's or 1000's of compounds with molecule structures and data using different dispensing methods to enable a large scale comparison, simply because it would require exposing confidential structures. To date there are only scatter plots on posters and in papers as we have referenced, and critically, none of these groups have reported the effect of molecular properties on these differences between dispensing methods.

Some of those other references are to posters and meeting presentations, so this seems to be one of those things that floats around in the field without landing explicitly in the literature. One of the paper's authors was good enough to send along the figure shown, which brings some of these data together, and it's an ugly sight. This paper is probably doing a real service in getting this potential problem out into the cite-able world: now there's something to point at.

How many other datasets are hosed up because of this effect? Now there's an important question, and one that we're not going to have an answer for any time soon. For some sets of compounds, there may be no problems at all, while others (as that graphic shows) can be a mess. There are, of course, plenty of projects where the assay numbers seem (more or less) to make sense, but there are plenty of others where they don't. Let the screener beware.

Update: here's a behind-the-scenes look at how this paper got published. It was not an easy path into the literature, by any means.

Second update: here's more about this at Nature Methods.

Comments (47) + TrackBacks (0) | Category: Drug Assays

April 16, 2013

What's Translational Synthesis, Anyway?

Email This Entry

Posted by Derek

There's another paper in the Nature Chemical Biology special issue that I wanted to mention, this one on "Translational Synthetic Chemistry". I can't say that I like the title, which seems to me to have a problem with reification (the process of trying to make something a thing which isn't necessarily a thing at all). I'm not so sure that there is a separate thing called "Translational Synthetic Chemistry", and I'm a bit worried that it might become a catch phrase all its own, which I think might lead to grief.

But that said, I still enjoyed the article. The authors are from H3 Biomedicine in Cambridge, which as I understand it is an offshoot of the Broad Institute and has several Schreiber-trained chemists on board. That means Diversity-Oriented Synthesis, of course, which is an area that I've expressed reservations about before. But the paper also discusses the use of natural product scaffolds as starting materials for new chemical libraries (a topic that's come up here and here), and the synthesis of diverse fragment collections beyond what we usually see. "Fragments versus DOS" has been set up before as a sort of cage match, but I don't think that has to be the case. And "Natural products versus DOS" has also been taken as a showdown, but I'm not so sure about that, either. These aren't either/or cases, and I don't think that the issues are illuminated by pretending that they are.

The authors end up calling for more new compound libraries, made by more new synthetic techniques, and assayed by newer and better high-throughput screens. Coming out against such recommendations makes a person feel as if they're standing up to make objections against motherhood and apple pies. And it's not that I think that these are bad ideas, but I just wonder if they're sufficient. Chemical space, as we were discussing the other day, is vast - crazily, incomprehensibly vast. Trying to blast off into it at random (which is what the pure DOS approaches have always seemed like to me) looks like something that a person could do for a century or two without seeing much return.

So if there are ways to increase the odds, I'm all for them. Natural-product-like molecules look like as good a way as any to do this, since they at least have the track record of evolution on their side. Things that are in roughly these kinds of chemical space, but which living organisms haven't gotten around to making, are still part of a wildly huge chemical space, but one that might have somewhat higher hit rates in screening. So Paul Hergenrother at Illinois might have the right idea when he uses natural products themselves as starting materials and makes new compound libraries from them.

So, who else is doing something like that? And what other methods do we have to make "natural-product-like" structures? Suggestions are welcome, and I'll assemble them and any ideas I have into another post.

Comments (53) + TrackBacks (0) | Category: Chemical News | Drug Assays

April 11, 2013

Fragments For Receptors: It Can Be Done

Email This Entry

Posted by Derek

The advent of real X-ray structures for receptors means that there are many experimental approaches that can now be tried that earlier would have been (most likely) foolhardy. My first research in the industry was on dopamine receptors, which I followed up by a stint on muscarinics, and we really did try to be rational drug designers. But that meant homology models and single-point mutations, and neither of those was always as helpful as you'd like. OK, fine: neither of them were very helpful at all, when you got right down to it. We kept trying to understand how our compounds were binding, but outside of the obvious GPCR features - gotta have a basic amine down there - we didn't get very far.

That's not to say that we didn't make potent, selective compounds. We certainly did, although you'll note that I'm not using the word "drug". For many of them, even the phrase "plausible clinical candidate" is difficult to get out with a straight face, potent and selective though they may have been. We made all these compounds, though, the old-fashioned way: straight SAR, add this on and take that away, fill out the table. Structural biology insights didn't really drive things much.

So when the transmembrane receptor X-ray structures began to show up, my first thought was whether or not they would have helped in that earlier effort, or whether they still had enough rough edges that they might have just helped to mislead us into thinking that we had things more figured out. There's a report, though, in the latest J. Med. Chem. that puts such structures to a pretty good test: can you use them to do fragment-based drug discovery?

Apparently so, at least up to the point described. This is the most complete example yet reproted of FBDD on a G-protein coupled receptor (beta-1 adrenergic). Given the prominence of receptors as drug targets, the late advent of fragment work in this field should tell you something about how important it is to have good structural information for a fragment campaign. I'm not sure if I've ever heard of one being successful without it - people say that it can be done, but I certainly wouldn't want to be the person doing it. That's not to say that X-ray structures are some sort of magic wand (this review should disabuse a person of that notion) - just that they're "necessary, but not sufficient" for getting a fragment program moving at reasonable speed. Otherwise, the amount of fumbling around at the edge of assay detection limits would be hard to take.

The beta-adrenergic receptor is the one with the most X-ray data available, with several different varieties of agonists and antagonists solved. So if any GPCR is going to get the fragment treatment, this would be the one. (There's also been a recent report of a fragment found for an adenosine receptor, which was largely arrived at through virtual screening). In this case, the initial screening was done via SPR (itself a very non-trivial technique for this sort of thing), followed by high-concentration radioligand assays, and eventual X-ray structure. They found a series of arylpiperazines, which are thoroughly believable as GPCR hits, although they don't have much of a history at the adrenergic receptor itself. The compounds are probably antagonists, mainly because they aren't making enough interactions to flip the switch to agonist, or not yet.

This paper only takes things up to this point, which is still a lot farther than anyone would have imagined a few years ago. My guess is that FBDD is still not ready for the spotlight in this field, though. This paper is from Miles Congreve and the folks at Heptares, world experts in GPCR crystallography, and presumably represents something pretty close to the state of the art. It's a proof-of-concept piece, but until the structures of more difficult receptors are available with more regularity, I don't think we'll see too much fragment work in the area. I'd be happy to be wrong about that.

Comments (7) + TrackBacks (0) | Category: Drug Assays

April 10, 2013

Pharmacology Versus Biology

Email This Entry

Posted by Derek

There's a comment made by CellBio to the recent post on phenotypic screening that I wanted to highlight I think it's an important point:

In drug discovery, we need fewer biologists dedicated to their biology, and more pharmacologists dedicated to testing the value of compounds.

He's not the first one to bemoan the decline of classic pharmacology. What we're talking about are the different answers to the (apparently) simple question, "What does this compound do?" The answer you're most likely to hear is something like "It's a such-and-such nanomolar Whateverase IV inhibitor". But the question could also be answered by saying what the compound does in cells, or in whole animals. It rarely is, though. We're so wedded to mechanisms and targets that we organize our thinking that way, and not always to our benefit.

In the case of the compound above, its cell activity may well be a consequence of its activity against Whateverase IV. If you have some idea of that protein's place in cellular processes, you might be fairly confident. But you can't really be sure about it. Do you have enzyme assays counterscreening against Whateverases I through V? How about the other enzymes with similar active sites? What, exactly, do all those things do in the cell if you are hitting them? Most of the time - all of the time, to be honest - we don't know the answers in enough detail.

So when people ask "What's this compound do?", what they really asking, most of the time, is "What does it do against the target that it was made for?" A better question would be "What does it do against everything it's been tested against?" But the most relevant questions for drug discovery are "What does it do to cells - and to animals? Or to humans?"

Update: Wavefunction, in the comments, mentions this famous article on the subject, which I was thinking of in the first paragraph after the quote above, but couldn't put my finger on. Thanks!

Comments (16) + TrackBacks (0) | Category: Drug Assays

April 9, 2013

Mass Spec Continues Its Conquests

Email This Entry

Posted by Derek

You know, mass spectrometry has been gradually taking over the world. Well, maybe not your world, but mine (and that of a lot of biopharma/biophysical researchers). There are just so many things that you can do with modern instrumentation that the assays and techniques just keep on coming.

This paper from a recent Angewandte Chemie is a good example. They're looking at post-translational modifications of proteins, which has always been a big field, and shows no signs of getting any smaller. The specific example here is SIRT1, an old friend to readers of this site, and the MALDI-based assay reported is a nice alternative to the fluorescence-based assays in that area, which have (notoriously) been shown to cause artifacts. The mass spec can directly detect deacetylation of a 16-mer histone H4 peptide - no labels needed.

The authors then screened a library of about 5500 natural product compounds (5 compounds per well in 384-well plates). As they showed, though, the hit rates observed would support higher pool numbers, and they successfully tested mixtures of up to 30 compounds at a time. Several structures were found to be micromolar inhibitors of the deacetylation reaction. None of these look very interesting or important per se, although some of them may find use as tool compounds. But the levels of detection and the throughput make me think that this might be a very useful technique for screening a fragment library.

Interestingly, they were also able to run the assay in the other direction, looking at acetylation of the histone protein, and discovered a new inhibitor of that process as well. These results prompted the authors to speculate that their assay conditions would be useful for a whole range of protein-modifying targets, and they may well be right.

So if this is such a good idea, why hasn't it been done before? The answer is that it has, especially if you go beyond the "open literature" and into the patents. Here, for example, is a 2009 application from Sirtris (who else?) on deacetylation/acetylation mass spec assays. And here's a paper (PDF) from 2009 (also in Angewandte) that used shorter peptides (6-mers) to profile enzymes of this type as well. There are many other assays of this sort that have been reported, or worked out inside various biopharma companies for their own uses. But this latest paper serves to show people (or remind them) that you can do such things on realistic substrates, with good reproducibility and throughput, and without having to think for a moment about coupled assays, scintillation plates, fluorescence windows, tagged proteins, and all the other typical details. Other things being equal, the more label-free your assay conditions, the better off you are. And other things are getting closer equal all the time.

Comments (3) + TrackBacks (0) | Category: Analytical Chemistry | Drug Assays

April 8, 2013

The Basics of Phenotypic Screening

Email This Entry

Posted by Derek

I wanted to mention another paper from Nature Chemical Biology's recent special issue, this one on the best ways to run phenotypic screens. This area has been making a comeback in recent years (as discussed around here before), so articles like this are very useful to help people get up to speed - similar to that article on fragment-based screening pitfalls I mentioned last week.

The author, Ulrike Eggert at King's College (London), has this observation about the cultural factors at work here:

Although my degrees are in chemistry, I have worked mostly in academic biology environments and have been immersed in the cultures of both disciplines. In my experience, biologists are very happy to use chemical tools if they are available (even if they are not optimal) but are less enthusiastic about engaging in small-molecule discovery. One of the reasons for this is that the academic funding culture in biology has focused almost entirely on hypothesis-driven research and has traditionally been dismissive of screening programs, which were considered to be nonintellectual fishing expeditions. With a growing appreciation for the value of interdisciplinary science and the serious need for new tools and approaches, this culture is slowly changing. Another reason is that some early phenotypic screens were perceived to have been only partial successes, resulting in 'low-quality' (for example, low-potency and nonselective) chemical probes.

These observations are right on target. The reaction of some academic biologists to screening programs reminds me of the reaction of some chemists to the "reaction discovery" schemes that have emerged in recent years: "Well, if you're just going to stagger around in circles until you trip over something, then sure. . ." But this, to me, just means that you should be careful to set up your discovery programs in the right places. One of my favorite quotes comes from Francis Crick, talking about the discovery of the double helix structure: "It's true that by blundering about we stumbled on gold, but the fact remains that we were looking for gold."

Eggert goes on to lay out the basic principles for success in this field. First, you'd better have clear, well-defined phenotypes as your readout, or you're sunk right from the start. Cell death is a pretty poor choice, for example, given the number of ways that you can kill a cell, and unfortunately, the same goes for inhibiting proliferation of cancer cells in vtiro. There really are an awful lot of compounds that will do that, in one cell line or another, and most of them are of no use at all. It's important to remember, though, that "well-defined" doesn't mean setting the boundaries so tight that you'll miss something interesting and unusual if it shows up - what it means is understanding your system well enough so that you'll recognize something unusual if it happens.

Assay design is, of course, critical. What's your signal-to-noise? How high is the throughput? How good are the positive and negative controls? What are the secondary assays that could be used to characterize your hits? And the point is also emphasized that the usual problem in these systems is not that you don't get any hits, but that you get so many that following them up is a problem all by itself. You're probably not going to find some compound that just lights up the assay perfectly all by itself - the more typical situation is a whole pile of different-looking things that might have worked, sort of. Sorting those out is a painful but essential part of the screen.

I'm a fan of phenotypic screening, personally, mainly because I don't think that we're smart enough to always realize what it is we're looking for, or exactly how to find it. But done suboptimally, this sort of screen is capable of wasting more time and effort than almost any other method. Eggert's article (and the references in it) are essential reading for anyone trying to get into the field. Medicinal chemists who find themselves working in this area for the first time should make sure to get caught up on these issues, because good med-chem followup is essential to any successful phenotypic campaign, and you want to make sure (as usual) that you're marching under the right flag.

Comments (7) + TrackBacks (0) | Category: Drug Assays

April 4, 2013

Good Ways To Mess Up Your Screening

Email This Entry

Posted by Derek

For those of you interested in fragment screening (and especially for those who are thinking of trying it out), Ben Davis of Vernalis and Dan Erlanson of Carmot Therapeutics have written an excellent guide to avoiding the common experimental problems. (Remarkably, at least to me, Elsevier has made this an open-access article). In fact, I'd recommend the article to everyone doing early-stage compound discovery, whether they're into fragments or not, because many of the issues it raises are universal.

. . .Clearly the approach works, but that is not to say it is easy. This Digest focuses on an area we believe is still insufficiently appreciated: the myriad pitfalls and artifacts that can befall a fragment-screening program. For the sake of brevity, we have chosen to focus on the problems that can hinder or derail an experimental fragment screening campaign; a full discussion of issues around fragment library design, virtual fragment screening, and fragment evolution is best dealt with elsewhere. . .

. . .Today many techniques are used to identify fragments, each with its own strengths. Importantly, however, each of these techniques also has unique limitations. While expert users are generally aware of these and readily pick out the signal from the noise, newcomers are often deceived by spurious signals. This can lead to resources wasted following up on artifacts. In the worst cases—unfortunately all too common—researchers may never realize that they have been chasing false positives, and publish their results. At best, this is an embarrassment, with the researchers sometimes none the wiser. At worst it can cause other research groups to waste their own resources. . .

They go into detail on difficulties with compound identity and stability (on storage and under the assay conditions). You've got your aggregators, your photoactive compounds, your redox cyclers, your hydrolytically unstable ones, etc., all of which can lead to your useless assay results and your wasted time. Then there's a discussion of the limits of each of the popular biophysical screening techniques (and they all have some), emphasizing that if you're going to do fragment screening, that you'd better be prepared to do more than one of these in every campaign. (They are, of course, quite right about this - I've seen the same sorts of situations that they report, where different assays yield different hit sets, and it's up to you to sort out which of those are real).

Highly recommended, as I say. These guys really know what they're talking about, and the drug discovery literature would be greatly improved if everyone were as well-informed.

Comments (7) + TrackBacks (0) | Category: Drug Assays

March 27, 2013

The DNA-Encoded Library Platform Yields A Hit

Email This Entry

Posted by Derek

I wrote here about DNA-barcoding of huge (massively, crazily huge) combichem libraries, a technology that apparently works, although one can think of a lot of reasons why it shouldn't. This is something that GlaxoSmithKline bought by acquiring Praecis some years ago, and there are others working in the same space.

For outsiders, the question has long been "What's come out of this work?" And there is now at least one answer, published in a place where one might not notice it: this paper in Prostaglandins and Other Lipid Mediators. It's not a journal whose contents I regularly scan. But this is a paper from GSK on a soluble epoxide hydrolase inhibitor, and therein one finds:

sEH inhibitors were identified by screening large libraries of drug-like molecules, each attached to a DNA “bar code”, utilizing DNA-encoded library technology [10] developed by Praecis Pharmaceuticals, now part of GlaxoSmithKline. The initial hits were then synthesized off of DNA, and hit-to-lead chemistry was carried out to identify key features of the sEH pharmacophore. The lead series were then optimized for potency at the target, selectivity and developability parameters such as aqueous solubility and oral bioavailability, resulting in GSK2256294A. . .

That's the sum of the med-chem in the article, which certainly compresses things, and I hope that we see a more complete writeup at some point from a chemistry perspective. Looking at the structure, though, this is a triaminotriazine-derived compound (as in the earlier work linked to in the first paragraph), so yes, you apparently can get interesting leads that way. How different this compound is from the screening hit is a good question, but it's noteworthy that a diaminotriazine's worth of its heritage is still present. Perhaps we'll eventually see the results of the later-generation chemistry (non-triazine).

Comments (12) + TrackBacks (0) | Category: Chemical Biology | Chemical News | Drug Assays | Drug Development

March 26, 2013

Automated Med-Chem, At Last?

Email This Entry

Posted by Derek

I've written several times about flow chemistry here, and a new paper in J. Med. Chem. prompts me to return to the subject. This, though, is the next stage in flow chemistry - more like flow med-chem:

Here, we report the application of a flow technology platform integrating the key elements of structure–activity relationship (SAR) generation to the discovery of novel Abl kinase inhibitors. The platform utilizes flow chemistry for rapid in-line synthesis, automated purification, and analysis coupled with bioassay. The combination of activity prediction using Random-Forest regression with chemical space sampling algorithms allows the construction of an activity model that refines itself after every iteration of synthesis and biological result.

Now, this is the point at which people start to get either excited or fearful. (I sometimes have trouble telling the difference, myself). We're talking about the entire early-stage optimization cycle here, and the vision is of someone topping up a bunch of solvent reservoirs, hitting a button, and leaving for the weekend in the expectation of finding a nanomolar compound waiting on Monday. I'll bet you could sell that to AstraZeneca for some serious cash, and to be fair, they're not the only ones who would bite, given a sufficiently impressive demo and slide deck.

But how close to this Lab of the Future does this work get? Digging into the paper, we have this:

Initially, this approach mirrors that of a traditional hit-to-lead program, namely, hit generation activities via, for example, high-throughput screening (HTS), other screening approaches, or prior art review. From this, the virtual chemical space of target molecules is constructed that defines the boundaries of an SAR heat map. An initial activity model is then built using data available from a screening campaign or the literature against the defined biological target. This model is used to decide which analogue is made during each iteration of synthesis and testing, and the model is updated after each individual compound assay to incorporate the new data. Typically the coupled design, synthesis, and assay times are 1–2 h per iteration.

Among the key things that already have to be in place, though, are reliable chemistry (fit to generate a wide range of structures) and some clue about where to start. Those are not givens, but they're certainly not impossible barriers, either. In this case, the team (three UK groups) is looking for BCL-Abl inhibitors, a perfectly reasonable test bed. A look through the literature suggested coupling hinge-binding motifs to DFG-loop binders through an acetylene linker, as in Ariad's ponatinib. This, while not a strategy that will earn you a big raise, is not one that's going to get you fired, either. Virtual screening around the structure, followed by eyeballing by real humans, narrowed down some possibilities for new structures. Further possibilities were suggested by looking at PDB structures of homologous binding sites and seeing what sorts of things bound to them.

So already, what we're looking at is less Automatic Lead Discovery than Automatic Patent Busting. But there's a place for that, too. Ten DFG pieces were synthesized, in Sonogashira-couplable form, and 27 hinge-binding motifs with alkynes on them were readied on the other end. Then they pressed the button and went home for the weekend. Well, not quite. They set things up to try two different optimization routines, once the compounds were synthesized, run through a column, and through the assay (all in flow). One will be familiar to anyone who's been in the drug industry for more than about five minutes, because it's called "Chase Potency". The other one, "Most Active Under Sampled", tries to even out the distributions of reactants by favoring the ones that haven't been used as often. (These strategies can also be mixed). In each case, the model was seeded with binding constants of literature structures, to get things going.

The first run, which took about 30 hours, used the "Under Sampled" algorithm to spit out 22 new compounds (there were six chemistry failures) and a corresponding SAR heat map. Another run was done with "Chase Potency" in place, generating 14 more compounds. That was followed by a combined-strategy run, which cranked out 28 more compounds (with 13 failures in synthesis). Overall, there were 90 loops through the process, producing 64 new products. The best of these were nanomolar or below.

But shouldn't they have been? The deck already has to be stacked to some degree for this technique to work at all in the present stage of development. Getting potent inhibitors from these sorts of starting points isn't impressive by itself. I think the main advantage to this is the time needed to generated the compound and the assay data. Having the synthesis, purification, and assay platform all right next to each other, with compound being pumped right from one to the other, is a much tighter loop than the usual drug discovery organization runs. The usual, if you haven't experienced it, is more like "Run the reaction. Work up the reaction. Run it through a column (or have the purification group run it through a column for you). Get your fractions. Evaporate them. Check the compound by LC/MS and NMR. Code it into the system and get it into a vial. Send it over to the assay folks for the weekly run. Wait a couple of days for the batch of data to be processed. Repeat."

The science-fictional extension of this is when we move to a wider variety of possible chemistries, and perhaps incorporate the modeling/docking into the loop as well, when it's trustworthy enough to do so. Now that would be something to see. You come back in a few days and find that the machine has unexpectedly veered off into photochemical 2+2 additions with a range of alkenes, because the Chase Potency module couldn't pass up a great cyclobutane hit that the modeling software predicted. And all while you were doing something else. And that something else, by this point, is. . .what, exactly? Food for thought.

Comments (16) + TrackBacks (0) | Category: Chemical News | Drug Assays | Drug Development

March 13, 2013

Getting Down to Protein-Protein Compounds

Email This Entry

Posted by Derek

Late last year I wrote about a paper that suggested that some "stapled peptides" might not work as well as advertised. I've been meaning to link to this C&E News article on the whole controversy - it's a fine overview of the area.

And that also gives me a chance to mention this review in Nature Chemistry (free full access). It's an excellent look at the entire topic of going after alpha-helix protein-protein interactions with small molecules. Articles like this really give you an appreciation for a good literature review - this information is scattered across the literature, and the authors here (from Leeds) have really done everyone interested in this topic a favor by collecting all of it and putting it into context.

As they say, you really have two choices if you're going after this sort of protein-protein interaction (well, three, if you count chucking the whole business and going to truck-driving school, but that option is not specific to this field). You can make something that's helical itself, so as to present the side chains in what you hope will be the correct orientation, or you can go after some completely different structure that just happens to arrange these groups into the right spots (but has no helical architecture itself).

Neither of these is going to lead to attractive molecules. The authors address this problem near the end of the paper, saying that we may be facing a choice here: make potent inhibitors of protein-protein interactions, or stay within Lipinski-guideline property space. Doing both at the same time just may not be possible. On the evidence so far, I think they're right. How we're going to get such things into cells, though, is a real problem (note this entry last fall on macrocyclic compounds, where the same concern naturally comes up). Since we don't seem to know much about why some compounds make it into cells and some don't, perhaps the way forward (for now) is to find a platform where as many big PPI candidates as possible can be evaluated quickly for activity (both in the relevant protein assay and then in cells). If we can't be smart enough, or not yet, maybe we can go after the problem with brute force.

With enough examples of success, we might be able to get a handle on what's happening. This means, though, that we'll have to generate a lot of complex structures quickly and in great variety, and if that's not a synthetic organic chemistry problem, I'd like to know what is. This is another example of a theme I come back to - that there are many issues in drug discovery that can only be answered by cutting-edge organic chemistry. We should be attacking these and making a case for how valuable the chemical component is, rather than letting ourselves be pigeonholed as a bunch of folks who run Suzuki couplings all day long and who might as well be outsourced to Fiji.

Comments (10) + TrackBacks (0) | Category: Drug Assays | Drug Development | Pharmacokinetics

March 4, 2013

Putting the (Hard) Chemistry Back in Med Chem

Email This Entry

Posted by Derek

While I'm on the subject of editorials, Takashi Tsukamoto of Johns Hopkins has one out in ACS Medicinal Chemistry Letters. Part of it is a follow-up to my own trumpet call in the journal last year (check the top of the charts here; the royalties are just flowing in like a river of gold, I can tell you). Tsukamoto is wondering, though, if we aren't exploring chemical space the way that we should:

One of the concerns is the likelihood of identifying drug-like ligands for a given therapeutic target, the so-called “druggability” of the target, has been defined by these compounds, representing a small section of drug-like chemical space. Are aminergic G protein coupled receptors (GPCRs) actually more druggable than other types of targets? Or are we simply overconcentrating on the area of chemical space which contains compounds likely to hit aminergic GPCRs? Is it impossible to disrupt protein–protein interactions with a small molecule? Or do we keep missing the yet unexplored chemical space for protein–protein interaction modulators because we continue making compounds similar to those already synthesized?

. . .If penicillin-binding proteins are presented as new therapeutic targets (without the knowledge of penicillin) today, we would have a slim chance of discovering β-lactams through our current medicinal chemistry practices. Penicillin-binding proteins would be unanimously considered as undruggable targets. I sometimes wonder how many other potentially significant therapeutic targets have been labeled as undruggable just because the chemical space representing their ligands has never been explored. . .

Good questions. I (and others) have had similar thoughts. And I'm always glad to see people pushing into under-represented chemical space (macrocycles being a good example).

The problem is, chemical space is large, and time (and money) are short. Given the pressures that research has been under, it's no surprise that everyone has been reaching for whatever will generate the most compounds in the shortest time - which trend, Tsukamoto notes, makes the whole med-chem enterprise that much easier to outsource to places with cheaper labor. (After all, if there's not so much skill involved in cranking out amides and palladium couplings, why not?)

My advice in the earlier editorial about giving employers something they can't buy in China and India still holds, but (as Tsukamoto says), maybe one of those things could (or should) be "complicated chemistry that makes unusual structures". Here's a similar perspective from Derek Tan at Sloan-Kettering, also referenced by Tsukamoto. It's an appealing thought, that we can save medicinal chemistry by getting back to medicinal chemistry. It may even be true. Let's hope so.

Comments (25) + TrackBacks (0) | Category: Chemical News | Drug Assays | Drug Industry History

February 27, 2013

Not What It Says On the Label, Though

Email This Entry

Posted by Derek

The topic of compound purity has come up here before, as well it should. Every experienced medicinal chemist knows that when you have an interesting new hit compound, that one of the first things to do is go back and make sure that it really is what it says on the label. Re-order it from the archive (in both powder and DMSO stock), re-order it if it's from a commercial source, and run it through the LC/MS and the NMR. (And as one of those links above says, if you have any thought that metal reagents were used to make the compound, check for those, too - they can be transparent to LC and NMR).

So when you do this, how many compounds flunk? Here are some interesting statistics from the folks at Emerald:

Recently, we selected a random set of commercial fragment compounds for analysis, and closely examined those that failed to better understand the reasons behind it. The most common reason for QC failure was insolubility (47%), followed by degradation or impurities (39%), and then spectral mismatch (17%) [Note: Compounds can acquire multiple QC designations, hence total incidences > 100% ]. Less than 4% of all compounds assayed failed due to solvent peak overlap or lack of non-exchangeable protons, both requirements for NMR screening. Failure rates were as high as 33% per individual vendor, with an overall average of 16%. . .

I very much wish that they'd identified that 33% failure rate vendor. But overall, they're suggesting that of 10 to 15% compounds will wipe out, regardless of source. Now, you may not feel that solubility is a key criterion for your work, because you're not doing NMR assays. (That's one that will only get worse as you move out of fragment-sized space, too). But that "degradation or impurities" category is still pretty significant. What are your estimates for commercial-crap-in-a-vial rates?

Comments (11) + TrackBacks (0) | Category: Chemical News | Drug Assays

Selective Inhibitor, The Catalog Says

Email This Entry

Posted by Derek

There's an interesting addendum to yesterday's post about natural product fragments. bAP15.pngDan Erlanson was pointing out that many of the proposed fragments were PAINS, and that prompted Jonathan Baell (author of the original PAINS paper) to leave a comment there mentioning this compound. Yep, you can buy that beast from Millipore, and it's being sold as a selective inhibitor of two particular enzymes. (Here's the original paper describing it). If it's really that selective, I will join one of those Greek monasteries where they eat raw onions and dry bread, and spend my time in atonement for ever thinking that a double nitrophenyl Schiff base enone with an acrylamide on it might be trouble.

Honestly, guys. Do a Ben Cravatt-style experiment across a proteome with that think, and see what you get. I'm not saying that it's going to absolutely label everything it comes across, but it's surely going to stick to more than two things, and have more effects than you can ascribe to those "selective" actions.

Comments (20) + TrackBacks (0) | Category: Chemical Biology | Drug Assays

February 26, 2013

Natural Product Fragments: Get Rid of the Ugly Ones Now

Email This Entry

Posted by Derek

Here's a paper at the intersection of two useful areas: natural products and fragments. Dan Erlanson over at Practical Fragments has a good, detailed look at the the work. What the authors have done is tried to break down known natural product structures into fragment-sized pieces, and cluster those together for guidance in assembling new screening libraries.

I'm sympathetic to that goal. I like fragment-based techniques, and I think that too many fragment libraries tend to be top-heavy with aromatic and heteroaromatic groups. Something with more polarity, more hydrogen-bonding character, and more three-dimensional structures would be useful, and natural products certainly fit that space. (Some of you may be familiar with a similar approach, the deCODE/Emerald "Fragments of Life", which Dan blogged about here). Synthetically, these fragments turn out to be a mixed bag, which is either a bug or a feature depending on your point of view (and what you have funding for or a mandate to pursue):

The natural-product-derived fragments are often far less complex structurally than the guiding natural products themselves. However, their synthesis will often still require considerable synthetic effort, and for widespread access to the full set of natural-product-derived fragments, the development of novel, efficient synthesis methodologies is required. However, the syntheses of natural-product-derived fragments will by no means have to meet the level of difficulty encountered in the multi-step synthesis of genuine natural products.

But take a look at Dan's post for the real downside:

Looking at the structures of some of the phosphatase inhibitors, however, I started to worry. One strong point of the paper is that it is very complete: the chemical structures of all 193 tested fragments are provided in the supplementary information. Unfortunately, the list contains some truly dreadful members; 17 of the worst are shown here, with the nasty bits shown in red. All of these are PAINS that will nonspecifically interfere with many different assays.

Boy, is he right about that, as you'll see when you take a look at the structures. They remind me of this beast, blogged about here back last fall. These structures should not be allowed into a fragment screening library; there are a lot of other things one could use instead, and their chances of leading only to heartbreak are just too high.

Comments (9) + TrackBacks (0) | Category: Chemical News | Drug Assays | Natural Products

February 21, 2013

The Hard Targets: How Far Along Are We?

Email This Entry

Posted by Derek

I wrote here about whole classes of potential drug targets that we really don't know how to deal with. It's been several years since then, and I don't think that the situation has improved all that much. (In 2011 I reviewed a book that advocated attacking these as a way forward for drug discovery).

Protein-protein interactions are still the biggest of these "undruggable targets", and there has been some progress made there. But I think we still don't have much in the way of general knowledge in this area. Every PPI target is its own beast, and you get your leads where you can, if you can. Transcription factors are the bridge between these and the protein-nucleic acid targets, which have been even harder to get a handle on (accounting for their appearance on lists like this one).

There are several chicken-and-egg questions in these areas. Getting chemical matter seems to be hard (that's something we can all agree on). Is that because we don't have compound collections that are biased the right way? If so, what the heck would the right way look like? Is is because we have trouble coming up with good screening techniques for some of these targets? (And if so, what are we lacking?) How much of the slower progress in these areas has been because of their intrinsic difficulty, and how much has been because people tend to avoid them (because of their, well, intrinsic difficulty?) If we all had our backs to the wall, could we do better, or would we generate just a lot more of the same?

I ask these questions because for years now, a lot of people in the industry have been saying that we need to get more of a handle on these things, because the good ol' small-molecule binding sites are getting scarcer. Am I right to think that we're still at the stage of telling each other this, or are there advances that I haven't kept up with?

Comments (14) + TrackBacks (0) | Category: Drug Assays | Drug Industry History

February 13, 2013

Mouse Models of Inflammation Are Basically Worthless. Now We Know.

Email This Entry

Posted by Derek

We go through a lot of mice in this business. They're generally the first animal that a potential drug runs up against: in almost every case, you dose mice to check pharmacokinetics (blood levels and duration), and many areas have key disease models that run in mice as well. That's because we know a lot about mouse genetics (compared to other animals), and we have a wide range of natural mutants, engineered gene-knockout animals (difficult or impossible to do with most other species), and chimeric strains with all sorts of human proteins substituted back in. I would not wish to hazard a guess as to how many types of mice have been developed in biomedical labs over the years; it is a large number representing a huge amount of effort.

But are mice always telling us the right thing? I've written about this problem before, and it certainly hasn't gone away. The key things to remember about any animal model is that (1) it's a model, and (2) it's in an animal. Not a human. But it can be surprisingly hard to keep these in mind, because there's no other way for a compound to become a drug other than going through the mice, rats, etc. No regulatory agency on Earth (OK, with the possible exception of North Korea) will let a compound through unless it's been through numerous well-controlled animal studies, for short- and long-term toxicity at the very least.

These thoughts are prompted by an interesting and alarming paper that's come out in PNAS: "Genomic responses in mouse models poorly mimic human inflammatory diseases". And that's the take-away right there, which is demonstrated comprehensively and with attention to detail.

Murine models have been extensively used in recent decades to identify and test drug candidates for subsequent human trials. However, few of these human trials have shown success. The success rate is even worse for those trials in the field of inflammation, a condition present in many human diseases. To date, there have been nearly 150 clinical trials testing candidate agents intended to block the inflammatory response in critically ill patients, and every one of these trials failed. Despite commentaries that question the merit of an overreliance of animal systems to model human immunology, in the absence of systematic evidence, investigators and public regulators assume that results from animal research reflect human disease. To date, there have been no studies to systematically evaluate, on a molecular basis, how well the murine clinical models mimic human inflammatory diseases in patients.

What this large multicenter team has found is that while various inflammation stresses (trauma, burns, endotoxins) in humans tend to go through pretty much the same pathways, the same is not true for mice. Not only do they show very different responses from humans (as measured by gene up- and down-regulation, among other things), they show different responses to each sort of stress. Humans and mice differ in what genes are called on, in their timing and duration of expression, and in what general pathways these gene products are found. Mice are completely inappropriate models for any study of human inflammation.

And there are a lot of potential reasons why this turns out to be so:

There are multiple considerations to our finding that transcriptional response in mouse models reflects human diseases so poorly, including the evolutional distance between mice and humans, the complexity of the human disease, the inbred nature of the mouse model, and often, the use of single mechanistic models. In addition, differences in cellular composition between mouse and human tissues can contribute to the differences seen in the molecular response. Additionally, the different temporal spans of recovery from disease between patients and mouse models are an inherent problem in the use of mouse models. Late events related to the clinical care of the patients (such as fluids, drugs, surgery, and life support) likely alter genomic responses that are not captured in murine models.

But even with all the variables inherent in the human data, our inflammation response seems to be remarkably coherent. It's just not what you see in mice. Mice have had different evolutionary pressures over the years than we have; their heterogeneous response to various sorts of stress is what's served them well, for whatever reasons.

There are several very large and ugly questions raised by this work. All of us who do biomedical research know that mice are not humans (nor are rats, nor are dogs, etc.) But, as mentioned above, it's easy to take this as a truism - sure, sure, knew that - because all our paths to human go through mice and the like. The New York Times article on this paper illustrates the sort of habits that you get into (emphasis below added):

The new study, which took 10 years and involved 39 researchers from across the country, began by studying white blood cells from hundreds of patients with severe burns, trauma or sepsis to see what genes are being used by white blood cells when responding to these danger signals.

The researchers found some interesting patterns and accumulated a large, rigorously collected data set that should help move the field forward, said Ronald W. Davis, a genomics expert at Stanford University and a lead author of the new paper. Some patterns seemed to predict who would survive and who would end up in intensive care, clinging to life and, often, dying.

The group had tried to publish its findings in several papers. One objection, Dr. Davis said, was that the researchers had not shown the same gene response had happened in mice.

“They were so used to doing mouse studies that they thought that was how you validate things,” he said. “They are so ingrained in trying to cure mice that they forget we are trying to cure humans.”

“That started us thinking,” he continued. “Is it the same in the mouse or not?”

What's more, the article says that this paper was rejected from Science and Nature, among other venues. And one of the lead authors says that the reviewers mostly seemed to be saying that the paper had to be wrong. They weren't sure where things had gone wrong, but a paper saying that murine models were just totally inappropriate had to be wrong somehow.

We need to stop being afraid of the obvious, if we can. "Mice aren't humans" is about as obvious a statement as you can get, but the limitations of animal models are taken so much for granted that we actually dislike being told that they're even worse than we thought. We aren't trying to cure mice. We aren't trying to make perfect diseases models and beautiful screening cascades. We aren't trying to perfectly match molecular targets with diseases, and targets with compounds. Not all the time, we aren't. We're trying to find therapies that work, and that goal doesn't always line up with those others. As painful as it is to admit.

Comments (50) + TrackBacks (0) | Category: Animal Testing | Biological News | Drug Assays | Infectious Diseases

February 12, 2013

The European Lead Factory

Email This Entry

Posted by Derek

That's what they're calling this new initiative, so points for bravery. I wrote about this proposal here last year, and now more details have emerged. The good news is that the former Merck facilities (well, Organon via Schering-Plough) in Newhouse (Scotland) and Oss (the Netherlands) will be part of the effort, so those will be put to some good use.

". . . the consortium consists of 30 academic and corporate partners, and aims to fill company pipelines with promising drug candidates. . .the initiative will build and curate a collection of 500,000 molecules for screening, 300,000 of which will come from the seven large pharmaceutical partners. The rest (are) intended to cover classes of biologically active molecule that are poorly represented in current libraries. . .Starting this July or August, the pharmaceutical partners will be able to use the library — including molecules from their competitors — in their own drug screens. Any academic group or company can also propose assays to test molecules in the library for biological activity."

Interestingly, both the compounds and the assay results will be proprietary (first refusal) for the people/organizations requesting them. The plan is for the whole thing to pay for itself through milestone payments and screening-for-hire for groups that are not part of the consortium. That's a worthy goal, but it's going to be complicated. One thing you can bet on is that the compounds in the collection themselves will not be the eventual ones that head to the clinic, so you get a "ship of Theseus" problem when you try to decide what belongs to whom (and at what point it started belonging). Note that the NIH's Molecular Libraries Program, by contrast, is made up of nonproprietary compounds, which were always assumed to be just starting points. (It's been having its own problems, too - many of the starting-point compounds, it seems, were - at least at first - probably never going to be starting points for anyone, so there still will not have been a clean head-to-head comparison between these two models).

And that brings up another difficulty with figuring out how things are going. I can certainly see why the results from the ELF will be proprietary, but that means that it may be some time before we can figure out whether it's providing anything worthwhile. The people running it will presumably have a better view.

Comments (8) + TrackBacks (0) | Category: Drug Assays

February 8, 2013

All Those Drug-Likeness Papers: A Bit Too Neat to be True?

Email This Entry

Posted by Derek

There's a fascinating paper out on the concept of "drug-likeness" that I think every medicinal chemist should have a look at. It would be hard to count the number of publications on this topic over the last ten years or so, but what if we've been kidding ourselves about some of the main points?

The big concept in this area is, of course, Lipinski criteria, or Rule of Five. Here's what the authors, Peter Kenny and Carlos Montanari of the University of São Paulo, have to say:

No discussion of drug-likeness would be complete without reference to the influential Rule of 5 (Ro5) which is essentially a statement of property distributions for compounds taken into Phase II clinical trials. The focus of Ro5 is oral absorption and the rule neither quantifies the risks of failure associated with non-compliance nor provides guidance as to how sub-optimal characteristics of compliant compounds might be improved. It also raises a number of questions. What is the physicochemical basis of Ro50s asymmetry with respect to hydrogen bond donors and acceptors? Why is calculated octanol/water partition coefficient (ClogP) used to specify Ro50s low polarity limit when the high polarity cut off is defined in terms of numbers of hydrogen bond donors and acceptors? It is possible that these characteristics reflect the relative inability of the octanol/water partitioning system to ‘see’ donors (Fig. 1) and the likelihood that acceptors (especially as defined for Ro5) are more common than donors in pharmaceutically-relevant compounds. The importance of Ro5 is that it raised awareness across the pharmaceutical industry about the relevance of physico- chemical properties. The wide acceptance of Ro5 provided other researchers with an incentive to publish analyses of their own data and those who have followed the drug discovery literature over the last decade or so will have become aware of a publication genre that can be described as ‘retrospective data analysis of large proprietary data sets’ or, more succinctly, as ‘Ro5 envy’.

There, fellow med-chemists, doesn't this already sound like something you want to read? Thought so. Here, have some more:

Despite widespread belief that control of fundamental physicochemical properties is important in pharmaceutical design, the correlations between these and ADMET properties may not actually be as strong as is often assumed. The mere existence of a trend is of no interest in drug discovery and strengths of trends must be known if decisions are to be accurately described as data-driven. Although data analysts frequently tout the statistical significance of the trends that their analysis has revealed, weak trends can be statistically significant without being remotely interesting. We might be confident that the coin that lands heads up for 51 % of a billion throws is biased but this knowledge provides little comfort for the person charged with predicting the result of the next throw. Weak trends can be beaten and when powered by enough data, even the feeblest of trends acquires statistical significance.

So, where are the authors going with all this entertaining invective? (Not that there's anything wrong with that; I'm the last person to complain). They're worried that the transformations that primary drug property data have undergone in the literature have tended to exaggerate the correlations between these properties and the endpoints that we care about. The end result is pernicious:

Correlation inflation becomes an issue when the results of data analysis are used to make real decisions. To restrict values of properties such as lipophilicity more stringently than is justified by trends in the data is to deny one’s own drug-hunting teams room to maneuver while yielding the initiative to hungrier, more agile competitors.

They illustrate this by reference to synthetic data sets, showing how one can get rather different impressions depending on how the numbers are handled along the way. Representing sets of empirical points by using their average values, for example, can cause the final correlations to appear more robust than they really are. That, the authors say, is just what happened in this study from 2006 ("Can we rationally design promiscuous drugs?) and in this one from 2007 ("The influence of drug-like concepts on decision-making in medicinal chemistry"). The complaint is that showing a correlation between cLogP and median compound promiscuity does not imply that there is one between cLogP and compound promiscuity per se. And the authors note that the two papers manage to come to opposite conclusions about the effect of molecular weight, which does make one wonder. The "Escape from flatland" paper from 2009 and the "ADMET rules of thumb" paper from 2008 (mentioned here) also come in for criticism on this point - binning averaged data from a large continuous set and then treated those as real objects for statistic analysis. Ones conclusions depend strongly on how many bins one uses. Here's a specific take on that last paper:

The end point of the G2008 analysis is ‘‘a set of simple interpretable ADMET rules of thumb’’ and it is instructive to examine these more closely. Two classifications (ClogP<4 and MW<400 Da; ClogP>4 or MW>400 Da) were created and these were combined with the four ionization state classifications to define eight classes of compound. Each combination of ADMET property and compound class was labeled according to whether the mean value of the ADMET property was lower than, higher than or not significantly different from the average for all compounds. Although the rules of thumb are indeed simple, it is not clear how useful they are in drug discovery. Firstly, the rules only say whether or not differences are significant and not how large they are. Secondly, the rules are irrelevant if the compounds of interest are all in the same class. Thirdly, the rules predict abrupt changes in ADMET properties going from one class to another. For example, the rules predict significantly different aqueous solubility for two neutral compounds with MW of 399 and 401 Da, provided that their ClogP values do not exceed 4. It is instructive to consider how the rules might have differed had values of logP and MW of 5 and 500 Da (or 3 and 300 Da) had been used to define them instead of 4 and 400 Da.

These problems also occur in graphical representations of all these data, as you'd imagine, and the authors show several of these that they object to. A particular example is this paper from 2010 ("Getting physical in drug discovery"). Three data sets, whose correlations in their primary data do not vary significantly, generate very different looking bar charts. And that leads to this comment:

Both the MR2009 and HY2010 studies note the simplicity of the relationships that the analysis has revealed. Given that drug discovery would appear to be anything but simple, the simplicity of a drug-likeness model could actually be taken as evidence for its irrelevance to drug discovery. The number of aromatic rings in a molecule can be reduced by eliminating rings or by eliminating aromaticity and the two cases appear to be treated as equivalent in both the MR2009 and HY2010 studies. Using the mnemonic suggested in MR2009 one might expect to make a compound more developable by replacing a benzene ring with cyclohexadiene or benzoquinone.

The authors wind up by emphasizing that they're not saying that things like lipophilicity, aromaticity, molecular weight and so on are unimportant - far from it. What they're saying, though, is that we need to be aware of how strong these correlations really are so that we don't fool ourselves into thinking that we're addressing our problems, when we really aren't. We might want to stop looking for huge, universally applicable sets of rules and take what we can get in smaller, local data sets within a given series of compounds. The paper ends with a set of recommendations for authors and editors - among them, always making primary data sets part of the supplementary material, not relying on purely graphical representations to make statistical points, and a number of more stringent criteria for evaluating data that have been partitioned into bins. They say that they hope that their paper "stimulates debate", and I think it should do just that. It's certainly given me a lot of things to think about!

Comments (14) + TrackBacks (0) | Category: Drug Assays | Drug Development | In Silico | The Scientific Literature

February 1, 2013

So How Does One Grow Beta-Cells?

Email This Entry

Posted by Derek

The short answer is "by looking for compounds that grow beta cells". That's the subject of this paper, a collaboration between Peter Schulz's group, the Novartis GNF. Schultz's group has already published on cell-based phenotypic screens in this area, where they're looking for compounds that could be useful in restoring islet function in patient with Type I diabetes.

These studies have used a rat beta-cell line (R7T1) that can be cultured, and they do good ol' phenotypic screening to look for compounds that induce proliferation (while not inducing it across the board in other cell types, of course). I'm a big fan of such approaches, but this is a good time to mention their limitations. You'll notice a couple of key words in that first sentence, namely "rat" and "cultured". Rat cells are not human cells, and cell lines that can be grown in vitro are not like primary cells from a living organism, either. If you base your entire approach this way, you run the risk of finding compounds that will, well, only work on rat cells in a dish. The key is to shift to the real thing as quickly as possible, to validate the whole idea.

That's what this paper does. The team has also developed an assay with primary human beta cells (which must be rather difficult to obtain), which are dispersed and plated. The tricky part seems to be keeping the plates from filling up with fibroblast cells, which are rather like the weeds of the cell culture world. In this case, their new lead compound (a rather leggy beast called WS-6) induced proliferation of both rat and human cells.

They took it on to an even more real-world system, mice that had been engineered to have a switchable defect in their own beta cells. Turning these animals diabetic, followed by treatment with the identified molecule (5 mpk, every other day), showed that it significantly lowered glucose levels compared to controls. And biopsies showed significantly increases beta-cell mass in the treated animals - all together, about as stringent a test as you can come up with in Type I studies.

So how does WS6 accomplish this? The paper goes further into affinity experiments with a biotinylated version of the molecule, which pulled down both the kinase IKK-epsilon and another target, Erb3 binding protein-1 (EBP1). An IKK inhibitor had no effect in the cell assay, interestingly, while siRNA experiments for EBP1 showed that knocking it down could induce proliferation. Doing both at the same time, though, had the most robust effect of all. The connection looks pretty solid.

Now, is WS6 a drug? Not at all - here's the conclusion of the paper:

In summary, we have identified a novel small molecule capable of inducing proliferation of pancreatic β cells. WS6 is among a few agents reported to cause proliferation of β cells in vitro or in vivo. While the extensive medicinal chemistry that would be required to improve the selectivity, efficacy, and tolerability of WS6 is beyond the scope of this work, further optimization of WS6 may lead to an agent capable of promoting β cell regeneration that could ultimately be a key component of combinatorial therapy for this complex disease.

Exactly so. This is excellent, high-quality academic research, and just the sort of thing I love to see. It tells us useful, actionable things that we didn't know about an important disease area, and it opens the door for a real drug discovery effort. You can't ask for more than that.

Comments (18) + TrackBacks (0) | Category: Chemical Biology | Diabetes and Obesity | Drug Assays

January 23, 2013

Eating A Whole Bunch of Random Compounds

Email This Entry

Posted by Derek

Reader Andy Breuninger, from completely outside the biopharma business, sends along what I think is an interesting question, and one that bears on a number of issues:

A question has been bugging me that I hope you might answer.

My understanding is that a lot of your work comes down to taking a seed molecule and exploring a range of derived molecules using various metrics and tests to estimate how likely they are to be useful drugs.

My question is this: if you took a normal seed molecule and a standard set of modifications, generated a set of derived molecules at random, and ate a reasonable dose of each, what would happen? Would 99% be horribly toxic? Would 99% have no effect? Would their effects be roughly the same or would one give you the hives, another nausea, and a third make your big toe hurt?

His impression of drug discovery is pretty accurate. It very often is just that: taking one or more lead compounds and running variations on them, trying to optimize potency, specificity, blood levels/absorption/clearance, toxicology, and so on. So, what do most of these compounds do in vivo?

My first thought is "Depends on where you start". There are several issues: (1) We tend to have a defined target in mind when we pick a lead compound, or (if it's a phenotypic assay that got us there), we have a defined activity that we've already seen. So things are biased right from the start; we're already looking at a higher chance of biological activity than you'd have by randomly picking something out of a catalog or drawing something on a board.

And the sort of target can make a big difference. There are an awful lot of kinase enzymes, for example, and compounds tend to cross-react with them, at least in the nearby families, unless you take a lot of care to keep that from happening. Compounds for the G-protein coupled biogenic amines receptors tend to do that, too. On the other hand, you have enzymes like the cytochromes and binding sites like the aryl hydrocarbon receptor - these things are evolved to recognize all sorts of structually disparate stuff. So against the right (or wrong!) sort of targets, you could expect to see a wide range of potential side activities, even before hitting the random ones.

(2) Some structural classes have a lot more biological activity than others. A lot of small-molecule drugs, for example, have some sort of basic amine in them. That's an important recognition element for naturally occurring substances, and we've found similar patterns in our own compounds. So something without nitrogens at all, I'd say, has a lower chance of being active in a living organism. (Barry Sharpless seems to agree with this). That's not to say that there aren't plenty of CHO compounds that can do you harm, just that there are proportionally more CHON ones that can.

Past that rough distinction, there are pharmacophores that tend to hit a lot, sometimes to the point that they're better avoided. Others are just the starting points for a lot of interesting and active compounds - piperazines and imidazoles are two cores that come to mind. I'd be willing to bet that a thousand random piperazines would hit more things than a thousand random morpholines (other things being roughly equal, like molecular weight and polarity), and either of them would hit a lot more than a thousand random cyclohexanes.

(3) Properties can make a big difference. The Lipinski Rule-of-Five criteria come in for a lot of bashing around here, but if I were forced to eat a thousand random compounds that fit those cutoffs, versus having the option to eat a thousand random ones that didn't, I sure know which ones I'd dig my spoon into.

And finally, (4): the dose makes the poison. If you go up enough in dose, it's safe to say that you're going to see an in vivo response to almost anything, including plenty of stuff at the supermarket. Similarly, I could almost certainly eat a microgram of any compound we have in our company's files with no ill effect, although I am not motivated to put that idea to the test. Same goes for the time that you're exposed. A lot of compounds are tolerated for single-dose tox but fail at two weeks. Compounds that make it through two weeks don't always make it to six months, and so on.

How closely you look makes the poison, too. We find that out all the time when we do animal studies - a compound that seems to cause no overt effects might be seen, on necropsy, to have affected some internal organs. And one that doesn't seem to have any visible signs on the tissues can still show effects in a full histopathology workup. The same goes for blood work and other analyses; the more you look, the more you'll see. If you get down to gene-chip analysis, looking at expression levels of thousands of proteins, then you'd find that most things at the supermarket would light up. Broccoli, horseradish, grapefruit, garlic and any number of other things would kick a full expression-profiling assay all over the place.

So, back to the question at hand. My thinking is that if you took a typical lead compound and dosed it at a reasonable level, along with a large set of analogs, then you'd probably find that if any of them had overt effects, they would probably have a similar profile (for good or bad) to whatever the most active compound was, just less of it. The others wouldn't be as potent at the target, or wouldn't reach the same blood levels. The chances of finding some noticeable but completely different activity would be lower, but very definitely non-zero, and would be wildly variable depending on the compound class. These effects might well cluster into the usual sorts of reactions that the body has to foreign substances - nausea, dizziness, headache, and the like. Overall, odds are that most of the compounds wouldn't show much, not being potent enough at any given target, or getting high enough blood levels to show something, but that's also highly variable. And if you looked closely enough, you'd probably find that that all did something, at some level.

Just in my own experience, I've seen one compound out of a series of dopamine receptor ligands suddenly turn up as a vasodilator, noticeable because of the "Rudolph the Red-Nosed Rodent" effect (red ears and tail, too). I've also seen compound series where they started crossing the blood-brain barrier more more effectively at some point, which led to a sharp demarcation in the tolerability studies. And I've seen many cases, when we've started looking at broader counterscreens, where the change of one particular functional group completely knocked a compound out of (or into) activity in some side assay. So you can never be sure. . .

Comments (22) + TrackBacks (0) | Category: Drug Assays | Drug Development | Pharma 101 | Pharmacokinetics | Toxicology

January 22, 2013

The Theology of Ligand Efficiency

Email This Entry

Posted by Derek

So in my post the other day about halogen bonds, I mentioned my unease at sticking in things like bromine and iodine atoms, because of the molecular weight penalty involved. Now, it's only a penalty if you're thinking in terms of ligand efficiency - potency per size of the molecule. I think that it's a very useful concept - one that was unheard of when I started in the industry, but which has now made a wide impression. The idea is that you should try, as much as possible, to make every part of your molecule worth something. Don't hang a chain off unless you're getting binding energy for it, and don't hang a big group off unless you're getting enough binding energy to make it worthwhile.

But how does one measure "worthwhile", or measure ligand efficiency in general? There are several schools of thought. One uses potency divided by molecular weight - there are different ways to make this come out to some sort of standard number, but that's the key operation. Another way, though, is to use potency divided by number of heavy atoms. These two scales will give you answers that are quite close to each other if you're just working in the upper reaches of the periodic table - there's not much difference between carbon, nitrogen, and oxygen. Sulfur will start throwing things off, as will chlorine But where the scales really give totally different answers, at least in common med-chem practice, is with bromine and iodine atoms. A single bromine (edit: fixed from earlier "iodine") weighs as much as a benzene ring, so the molecular-weight-based calculation takes a torpedo, while the heavy atom count just registers one more of the things.

For that very reason, I've been in the molecular-weight camp. But TeddyZ of Practical Fragments showed up in the comments to the halogen bond post, recommending arguments for the other side. But now that I've checked those out, I'm afraid that I still don't find them very convincing.

That's because the post he's referring to makes the case against simple molecular weight cutoffs alone. I'm fine with that. There's no way that you can slice things up by a few mass units here and there in any meaningful way. But the issue here isn't just molecular weight, it's activity divided by weight, and in all the cases shown, the ligand efficiency for the targets of these compounds would have gone to pieces if the "smaller" analog were picked. From a ligand efficiency standpoint, these examples are straw men.

So I still worry about bromine and iodine. I think that they hurt a compound's properties, and that treating them as "one heavy atom", as if they were nitrogens, ignores that. Now, that halogen bond business can, in some cases, make up for that, but medicinal chemists should realize the tradeoffs they're making, in this case as in all the others. I wouldn't, for example, rule out an iodo compound as a drug candidate, just because it's an iodo compound. But that iodine had better be earning its keep (and probably would be doing so via a halogen bond). It has a lot to earn back, too, considering the possible effects on PK and compound stability. Those would be the first things I would check in detail if my iodo candidate led the list in the other factors, like potency and selectivity. Then I'd get it into tox as soon as possible - I have no feel whatsoever for how iodine-substituted compounds act in whole-animal tox studies, and I'd want to find out in short order. That, in fact, is my reaction to unusual structures of many kinds. Don't rule them out a priori; but get to the posteriori part, where you have data, as quickly as possible.

So, thoughts on heavy atoms? Are there other arguments to make in favor of ligand efficiency calculated that way, or do most people use molecule weight?

Comments (26) + TrackBacks (0) | Category: Drug Assays | Drug Development

January 10, 2013

Automated Ligand Design?

Email This Entry

Posted by Derek

There's a paper out in Nature with the provocative title of "Automated Design of Ligands to Polypharmcological Profiles". Admittedly, to someone outside my own field of medicinal chemistry, that probably sounds about as dry as the Atacama desert, but it got my attention.

It's a large multi-center contribution, but the principal authors are Andrew Hopkins at Dundee and Bryan Roth at UNC-Chapel Hill. Using James Black's principle that the best place to find a new drug is to start with an old drug, what they're doing here is taking known ligands and running through a machine-learning process to see if they can introduce new activities into them. Now, those of us who spend time trying to take out other activities might wonder what good this is, but there are a some good reasons: for one thing, many CNS agents are polypharmacological to start with. And there certainly are situations where you want dual-acting compounds, CNS or not, which can be a major challenge. And read on - you can run things to get selectivity, too.

So how well does their technique work? The example they give starts with the cholinesterase inhibitor donepezil (sold as Aricept), which has a perfectly reasonable med-chem look to its structure. The groups' prediction, using their current models, was the it had a reasonable chance of having D4 dopaminergic activity, but probably not D2 (which numbers were borne out by experiment, and might have something to do with whatever activity Aricept has for Alzheimer's). I'll let them describe the process:

We tested our method by evolving the structure of donepezil with the dual objectives of improving D2 activity and achieving blood–brain barrier penetration. In our approach the desired multi-objective profile is defined a priori and then expressed as a point in multi-dimensional space termed ‘the ideal achievement point’. In this first example the objectives were simply defined as two target properties and therefore the space has two dimensions. Each dimension is defined by a Bayesian score for the predicted activity and a combined score that describes the absorption, distribution, metabolism and excretion (ADME) properties suitable for blood–brain barrier penetration (D2 score = 100, ADME score = 50). We then generated alternative chemical structures by a set of structural transformations using donepezil as the starting structure. The population was subsequently enumerated by applying a set of transformations to the parent compound(s) of each generation. In contrast to rules-based or synthetic-reaction-based approaches for generating chemical structures, we used a knowledge-based approach by mining the medicinal chemistry literature. By deriving structural transformations from medicinal chemistry, we attempted to mimic the creative design process.

Hmm. They rank these compounds in multi-dimensional space, according to distance from the ideal end point, filter them for chemical novelty, Lipinski criteria, etc., and then use the best structures as starting points for another round. This continues until you reach close enough to the desired point, or until you dead-end on improvement. In this case, they ended up with fairly active D2 compounds, by going to a lactam in the five-membered ring, lengthening the chain a bit, and going to an arylpiperazine on the end. They also predicted, though, that these compounds would hit a number of other targets, which they indeed did on testing.

How about something a bit more. . .targeted? They tried taking these new compounds through another design loop, this time trying to get rid of all the alpha-adrenergic activity they'd picked up, while maintaining the 5-HT1A and dopamine receptor activity they now had. They tried it both ways - running the algorithms with filtration of the alpha-active compounds at each stage, and without. Interestingly, both optimizations came up with very similar compounds, differing only out on the arylpiperazine end. The alpha-active series wanted ortho-methoxyphenyl on the piperazine, while the alpha-inactive series wanted 2-pyridyl. These preferences were confirmed by experiment as well. Some of you who've worked on adrenergics might be saying "Well, yeah, that's what the receptors are already known to prefer, so what's the news here?" But keep in mind, what the receptors are known to prefer is what's been programmed into this process, so of course, that's what it's going to recapitulate. The idea is for the program to keep track of all the known activities - the huge potential SAR spreadsheet - so you don't have to try to do it yourself, with you own grey matter.

The last example asks whether, starting from donezepil, potent and selective D4 compounds could be evolved. I'm going to reproduce the figure from the paper here, to give an idea of the synthetic transformations involved:
So, donezepil (compound 1) is 614 nM against D4, and after a few rounds of optimization, you get structure 13, which is 9 nM. Not bad! Then if you take 13 as a starting point, and select for structural novelty along the way, you get 18 (five micromolar against D4), 20, 21, and (S)-27 (which is 90 nM at D4). All of these compounds picked up a great deal more selectivity for D4 compared to the earlier donezepil-derived scaffolds as well.

Well, then, are we all out of what jobs we have left? Not just yet. You'll note that the group picked GPCRs as a field to work in, partly because there's a tremendous amount known about their SAR preferences and cross-functional selectivities. And even so, of the 800 predictions made in the course of this work, the authors claim about a 75% success rate - pretty impressive, but not the All-Seeing Eye, quite yet. I'd be quite interested in seeing these algorithms tried out on kinase inhibitors, another area with a wealth of such data. But if you're dwelling among the untrodden ways, like Wordsworth's Lucy, then you're pretty much on your own, I'd say, unless you 're looking to add in some activity in one of the more well-worked-out classes.

But knowledge piles up, doesn't it? This approach is the sort of thing that will not be going away, and should be getting more powerful and useful as time goes on. I have no trouble picturing an eventual future where such algorithms do a lot of the grunt work of drug discovery, but I don't foresee that happened for a while yet. Unless, of course, you do GPCR ligand drug discovery. In that case, I'd be contacting the authors of this paper as soon as possible, because this looks like something you need to be aware of.

Comments (12) + TrackBacks (0) | Category: Drug Assays | In Silico | The Central Nervous System

January 8, 2013

Metal Impurities Will Waste Your Time

Email This Entry

Posted by Derek

Here's a paper on a high-throughput screening issue that everyone in the field should be aware of: metal impurities in your compounds. The group (from Roche) describes a recent experience, and I think that many readers will shiver in recognition:

The hits were resynthesized, and close analogues were prepared for early structure−activity relationship (SAR) exploration. All three series lacked conclusive SAR. Most exemplifying are the activities of different batches of the very same compounds that exhibited very different activities from being low micromolar to inactive with IC50 values greater than one millimolar (Table 1). Additionally, the SAR of close analogues was either flat or very steep as indicated by compounds with minimal structural changes losing all activity (data not shown).

For these particular hits, we investigated these findings further. It was discovered that for one series, different routes of synthesis were used for the original preparation of the HTS library compound and its resynthesis. The historic synthesis made use of a zinc/titanium reduction step, whereas the new synthesis leading to inactive compounds did not. The schemes to prepare compounds of the other series also had steps involving zinc. Elemental analysis of the samples to determine the zinc content revealed that the active batches contained different amounts of zinc of up to 20% of total mass, whereas the inactive batches only had traces. . .

I think that many of us have been burned by this in the past, but it's something that should be out there in the literature so that it's easier to make the case to those who haven't heard about it. The Roche group suggests a counterscreen using a zinc chelator (TPEN) that will get rid of zinc-based effects. They pulled out 90 of their hits based on that work, and checking those against past assays showed that they had unusually high hit rates across the years. Some of them had, in fact, been considered hits for current Roche projects, and checking those assays showed that they were sensitive to zinc as well.

I can tell you from personal experience that the stuff can be a real problem. In that case, "impurity" was a relative term - the compound from the files turned out to be a 1:1 zinc complex, not that this little fact was noted anywhere in its (rather ancient) documentation from inside the company.

And I've seen copper do the same sort of thing. I would very much recommend checking out any active compound that looks to have been made by Ullmann chemistry or the like. I mean, I like the Ullmann reaction (and it looks like I may be setting some of them up soon), but there's a lot of copper in those things, and some assays are very sensitive to it. In extreme cases, I've seen compounds come in from custom synthesis houses that were colored green from the stuff, and that's just not going to work out. There are regrettably few lead-like compounds that come by a green tint honestly: you're looking at copper, maybe chromium, or perhaps even nickel, none of which will help you generate reliable assay numbers. Don't even let the green stuff into your collection, if you can - clean it up first, and complain to the people who sent it to you. (Note, by contrast, that zinc complexes tend to show no added colors).

Jonathan Beall speculated to me in an e-mail that maybe this is one way that frequent-hitter compounds can get on such lists, by coordinating metals. It's certainly possible. Ignore metals at your peril!

Comments (23) + TrackBacks (0) | Category: Drug Assays

January 2, 2013

How Many Good Screening Compounds Are There?

Email This Entry

Posted by Derek

So, how many good screening compounds are there to be had? We can now start to argue about the definition of "good"; that's the traditional next step in this process. But there's a new paper from Australia's Jonathan Baell on this very question that's worth a look.

He and his co-workers have already called attention to the number of compounds with possibly problematic functional groups for high-throughput screening. In this paper, he also quantifies the way that commercial compound collections tend to go wild on certain scaffolds - giving you, say, three hundred of one series and one of another. One does not mind diagnosing synthetic ease as the reason for this. And it's not always bad - if you get a hit from the series, then you have an SAR collection ready to go in the follow-up. But you wouldn't necessarily want all of them in there for the first go-round.

But there are many other criteria, and as anyone who's done the exercise can appreciate, large lists of compounds tend to be cut down to size rapidly. The paper shows this in action with a commercial set of 400,000 compounds. Apply some not-too-stringent criteria (between 1 and 4 rings, molecular weights between 150 and 450, cLogP less than 6, no more than 5 hydrogen bond donors and no more than 8 acceptors, up to three chiral centers, and up to 12 rotatable bonds), and you're down to 250K compounds right there. Clear out some functional groups and the PAINS list, and you're down to 170K. Want to cut the molecular weight down to 400, and rotatable bonds down to 10? 130,000 compounds remain. cLogP only up to 5, donors down to 3 or fewer, acceptors down to 6 or fewer? 110,000.

At this point, the paper says, further inspection of the list led to the realization that there were still a lot of problematic functional groups present. (I had a similar experience myself recently, filtered down a less humungous list. Even after several rounds, I was surprised to find, on looking more closely, how many oximes, hydrazones, Schiff bases, hydrazines, and N-hydroxyls were left). In Baell's case, clearing out the not-so-great at this point cut things down to 50,000 compounds. Then a Tanimoto cutoff (to get rid of things that were at least 90% similar to the existing screening compounds) cleared out all but 10,000. Applying the same cutoff, but getting rid of compounds on the list that were more than 90% similar to each other, reduced it to 6,000. So, in other words, one could make a good case for getting rid of over 98% of the vendor's list for high-throughput screening purposes. Similar results were obtained for many other commercial sets of compounds; the paper has the exact numbers (although not, alas, the vendor names involved!)

There were other vendor considerations as well. By the time Baell and his group had gone through all this compound-crunching and placed orders, significant numbers of compounds turned out to be unavailable. (I'm willing to bet that quite a few of them would have turned out to be unavailable even if they'd placed their orders that afternoon, but I'm of a cynical bent). That catalog turnover also brings up the problem of being able to re-order compounds if they turn out to be hits:

. . .there were only two vendors whose resupply philosophy we considered to be sound, this philosophy being that around 40 mg stock was set aside and accessible exclusively to prior buyers of that compound for the purposes of resupply of ca. 1−2 mg for secondary assay of a confirmed screening hit. We believe this issue of resupply is in urgent need of attention by vendors and will provide a competitive edge to those vendors willing to better guarantee resupply.

By the time they'd surveyed the various large-scale compound vendors, the group had looked over the majority of commercially available screening compounds. Given the attrition rates, how many actual compounds would cover the world's purchasable chemical space? The best guess is about 340,000, of the many millions of potentially purchasable items.

Of course, all these numbers are subject to dispute - you may not agree with some of the functional group or property cutoffs, or you might want things cut down even more. The paper addresses this question, and the general one of why any particular compound should be in a screening collection at all. My own criterion is "Would I be willing to follow up on this compound if it were a hit?" But different chemists, as has been proven many times, will answer such questions in different ways.

A big part of the discussion are those Tanimoto similarity scores, and the paper has a good deal to say about that. You wouldn't want to cut everything down to just singleton compounds (most likely), but you also don't need to have dozens and dozens of para-chloro/para-fluoro methyl-ethyl analogs in each series, either. The best guess is that most vendor catalogs are still rather unbalanced: they have far too many analogs for some compound classes, but too few for many more. Singleton compounds represent most of the chemical diversity for many collections, but you could make the case that there shouldn't be any singletons, ideally. Even two or three representatives from each structural class would be a real improvement. A vendor collection of 400,000 compounds that consisted of 40,000 fairly distinct structures with ten members of each class would be something to see - but no one's ever seen such a thing.

This new paper, by the way, is full of references to the screening-collection literature, as well as discussing many of the issues itself. I recommend it to anyone thinking about these issues; there are a lot of things that you don't want to have to rediscover!

Comments (17) + TrackBacks (0) | Category: Drug Assays

December 6, 2012

Four Million Compounds to Screen

Email This Entry

Posted by Derek

There's a new paper out that does something unique: it compares the screening libraries of two large drug companies, both of which agreed to open their books to each other (up to a point) for the exercise. The closest analog that I know of is when Bayer merged with/bought Schering AG, and the companies published on the differences between the two compound collections as they worked on merging them. (As a sideline, I hope that they've culled some of the things that were in that collection when I worked there. I actually had a gallery of horrible compounds from the files that I kept around to amaze people - it was hard to come up with a functional group that wasn't represented somewhere). That combined Bayer collection 2.75 million compounds) has now been compared with AstraZeneca's (1.4 million compounds). The two of them have clearly been exploring precompetitive collaboration in high-throughput screening, and trying to figure out how much there is to gain.

The first question that comes to mind is how the companies managed this - after all, you wouldn't want another outfit to actually stroll through your structures. They used 2-D fingerprints to get around this problem, the ECFP4 system, to be exact. That's a descriptor that gives a lot of structural information without being reversible; you can't reassemble the actual compound from the fingerprint.

So what's in these collections, and how much do the two overlap? I think that the main take-away from the paper is the answer to the second question, which is "Not as much as you'd think". Using Tanimoto similarity calculations (ratio of the intersecting set to the union set) for all those molecular fingerprints (with a cutoff of 0.70 for "similar"), they found that about 144,000 compounds in the Bayer collection seem to be duplicated in the AstraZeneca collection. Not surprisingly, these turned out to be commercially available; they'd been bought from the same vendors, most likely. That's not much!

Considering that all pharmaceutical companies can access the same external vendors this number is certainly lower than expected. There are 290K compounds that are not identical but very similar between both databases, with nearest neighbors with Tanimoto values in the range of 0.7–1.0. In a joint HTS campaign this would lead to a higher coverage of the chemical space in SAR exploration. The remaining 2.3M compounds of the Bayer collection have no similar compounds in the AstraZeneca collection, as is reflected in nearest neighbors with Tanimoto values ≤0.7. Thus, a practical interpretation is that AstraZeneca would extend their available chemical space with 2.3M novel, distinct chemical entities by testing the Bayer Pharma AG collection in a HTS campaign, provided that intellectual property issues could be resolved.

One interesting effect, though, is that compounds which would be classed as "singletons" in each collection (and thus could be a bit problematic to follow up on) had closer relatives over in the other company's collection. That could be a real advantage, rescuing what might otherwise be a collection of unrelated stuff - a few legitimate leads buried in a bunch of tedious compounds that would eventually have to be discarded one by one.

The teams also compared their collections to a large public on, the ChEMBL database:

The public ChEMBL database was chosen to simulate a third-party compound collection. It consisted of 600K molecules derived from medicinal chemistry publications annotated with pharmacological/biological data. Hence, we used this source as a proxy for ‘a pharmaceutical’ compound collection. We opted to avoid the use of commercial screening collections for this assessment as it would clearly reveal the number and source of acquisitions. In Fig. 6, we display the distribution of the nearest neighbors in the ChEMBL compounds (query collection) to the target collection corresponding to the merged AstraZeneca and Bayer Pharma AG compounds. Despite the huge set of more than 3.7 million compounds to which the relatively small ChEMBL collection is compared, more than 80% of this collection has their nearest neighbor with a Tanimoto index below 0.70. Consistent with the volume of published and patented compounds this result again emphasize that even in large collections there is still relevant unexplored chemical space accessed by other groups in industry and academia.

So the question comes up, after all these comparisons: have the two companies decided to do anything about this? The conclusions of the paper seem clear. If you're interested in high-throughput screening, combining the two collections would significantly improve the results obtained from screening either one alone. How much value does either company assign to that, compared to the intellectual property risks involved? The decision (or lack of decision) that's reached on this will serve as the best answer: revealed preference always wins out over stated preference.

Comments (15) + TrackBacks (0) | Category: Drug Assays

November 30, 2012

A Broadside Against The Way We Do Things Now

Email This Entry

Posted by Derek

There's a paper out in Drug Discovery Today with the title "Is Poor Research the Cause of Declining Productivity in the Drug Industry? After reviewing the literature on phenotypic versus target-based drug discovery, the author (Frank Sams-Dodd) asks (and has asked before):

The consensus of these studies is that drug discovery based on the target-based approach is less likely to result in an approved drug compared to projects based on the physiological- based approach. However, from a theoretical and scientific perspective, the target-based approach appears sound, so why is it not more successful?

He makes the points that the target-based approach has the advantages of (1) seeming more rational and scientific to its practitioners, especially in light of the advances in molecular biology over the last 25 years, and (2) seeming more rational and scientific to the investors:

". . .it presents drug discovery as a rational, systematic process, where the researcher is in charge and where it is possible to screen thousands of compounds every week. It gives the image of industrialisation of applied medical research. By contrast, the physiology-based approach is based on the screening of compounds in often rather complex systems with a low throughput and without a specific theory on how the drugs should act. In a commercial enterprise with investors and share-holders demanding a fast return on investment it is natural that the drug discovery efforts will drift towards the target-based approach, because it is so much easier to explain the process to others and because it is possible to make nice diagrams of the large numbers of compounds being screened.

This is the "Brute Force bias". And he goes on to another key observation: that this industrialization (or apparent industrialization) meant that there were a number of processes that could be (in theory) optimized. Anyone who's been close to a business degree knows how dear process optimization is to the heart of many management theorists, consultants, and so on. And there's something to that, if you're talking about a defined process like, say, assembling pickup trucks or packaging cat litter. This is where your six-sigma folks come in, your Pareto analysis, your Continuous Improvement people, and all the others. All these things are predicated on the idea that there is a Process out there.

See if this might sound familiar to anyone:

". . .the drug dis- covery paradigm used by the pharmaceutical industry changed from a disease-focus to a process-focus, that is, the implementation and organisation of the drug discovery process. This meant that process-arguments became very important, often to the point where they had priority over scientific considerations, and in many companies it became a requirement that projects could conform to this process to be accepted. Therefore, what started as a very sensible approach to drug discovery ended up becoming the requirement that all drug dis- covery programmes had to conform to this approach – independently of whether or not sufficient information was available to select a good target. This led to dogmatic approaches to drug discovery and a culture developed, where new projects must be presented in a certain manner, that is, the target, mode-of-action, tar- get-validation and screening cascade, and where the clinical manifestation of the disease and the biological basis of the disease at systems-level, that is, the entire organism, were deliberately left out of the process, because of its complexity and variability.

But are we asking too much when we declare that our drugs need to work through single defined targets? Beyond that, are we even asking too much when we declare that we need to understand the details of how they work at all? Many of you will have had such thoughts (and they've been expressed around here as well), but they can tend to sound heretical, especially that second one. But that gets to the real issue, the uncomfortable, foot-shuffling, rather-think-about-something-else question: are we trying to understand things, or are we trying to find drugs?

"False dichotomy!", I can hear people shouting. "We're trying to do both! Understanding how things work is the best way to find drugs!" In the abstract, I agree. But given the amount there is to understand, I think we need to be open to pushing ahead with things that look valuable, even if we're not sure why they do what they do. There were, after all, plenty of drugs discovered in just that fashion. A relentless target-based environment, though, keeps you from finding these things at all.

What it does do, though, is provide vast opportunities for keeping everyone busy. And not just "busy" in the sense of working on trivia, either: working out biological mechanisms is very, very hard, and in no area (despite decades of beavering away) can we say we've reached the end and achieved anything like a complete picture. There are plenty of areas that can and will soak up all the time and effort you can throw at them, and yield precious little in the way of drugs at the end of it. But everyone was working hard, doing good science, and doing what looked like the right thing.

This new paper spends quite a bit of time on the mode-of-action question. It makes the point that understanding the MoA is something that we've imposed on drug discovery, not an intrinsic part of it. I've gotten some funny looks over the years when I've told people that there is no FDA requirement for details of a drug's mechanism. I'm sure it helps, but in the end, it's efficacy and safety that carry the day, and both of those are determined empirically: did the people in the clinical trials get better, or worse?

And as for those times when we do have mode-of-action information, well, here are some fighting words for you:

". . .the ‘evidence’ usually involves schematic drawings and flow-diagrams of receptor complexes involving the target. How- ever, it is almost never understood how changes at the receptor or cellular level affect the phy- siology of the organism or interfere with the actual disease process. Also, interactions between components at the receptor level are known to be exceedingly complex, but a simple set of diagrams and arrows are often accepted as validation for the target and its role in disease treatment even though the true interactions are never understood. What this in real life boils down to is that we for almost all drug discovery programmes only have minimal insight into the mode-of-action of a drug and the biological basis of a disease, meaning that our choices are essentially pure guess-work.

I might add at this point that the emphasis on defined targets and mode of action has been so much a part of drug discovery in recent times that it's convinced many outside observers that target ID is really all there is to it. Finding and defining the molecular target is seen as the key step in the whole process; everything past that is just some minor engineering (and marketing, naturally). That fact that this point of view is a load of fertilizer has not slowed it down much.

I think that if one were to extract a key section from this whole paper, though, this one would be a good candidate:

". . .it is not the target-based approach itself that is flawed, but that the focus has shifted from disease to process. This has given the target-based approach a dogmatic status such that the steps of the validation process are often conducted in a highly ritualised manner without proper scientific analysis and questioning whether the target-based approach is optimal for the project in question.

That's one of those "Don't take this in the wrong way, but. . ." statements, which are, naturally, always going to be taken in just that wrong way. But how many people can deny that there's something to it? Almost no one denies that there's something not quite right, with plenty of room for improvement.

What Sams-Dodd has in mind for improvement is a shift towards looking at diseases, rather than targets or mechanisms. For many people, that's going to be one of those "Speak English, man!" moments, because for them, finding targets is looking at diseases. But that's not necessarily so. We would have to turn some things on their heads a bit, though:

In recent years there have been considerable advances in the use of automated processes for cell-culture work, automated imaging systems for in vivo models and complex cellular systems, among others, and these developments are making it increasingly possible to combine the process-strengths of the target-based approach with the disease-focus of the physiology-based approach, but again these technologies must be adapted to the research question, not the other way around.

One big question is whether the investors funding our work will put up with such a change, or with such an environment even if we did establish it. And that gets back to the discussion of Andrew Lo's securitization idea, the talk around here about private versus public financing, and many other topics. Those I'll reserve for another post. . .

Comments (30) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History | Who Discovers and Why

November 29, 2012

Roche Repurposes

Email This Entry

Posted by Derek

Another drug repurposing initiative is underway, this one between Roche and the Broad Institute. The company is providing 300 failed clinical candidates to be run through new assays, in the hopes of finding a use for them.

I hope something falls out of this, because any such compounds will naturally have a substantial edge in further development. They should all have been through toxicity testing, they've had some formulations work done on them, a decent scale-up route has been identified, and so on. And many of these candidates fell out in Phase II, so they've even been in human pharmacokinetics.

On the other hand (there's always another hand), you could also say that this is just another set of 300 plausible-looking compounds, and what does a 300-compound screening set get you? The counterargument to this is that these structures have not only been shown to have good absorption and distribution properties (no small thing!), they've also been shown to bind well to at least one target, which means that they may well be capable of binding well to other similar motifs in other active sites. But the counterargument to that is that now you've removed some of those advantages in the paragraph above, because any hits will now come with selectivity worries, since they come with guaranteed activity against something else.

This means that the best case for any repurposed compound is for its original target to be good for something unanticipated. So that Roche collection of compounds might also be thought of as a collection of failed targets, although I doubt if there are a full 300 of those in there. Short of that, every repurposing attempt is going to come with its own issues. It's not that I think these shouldn't be tried - why not, as long as it doesn't cost too much - but things could quickly get more complicated than they might have seemed. And that's a feeling that any drug discovery researcher will recognize like an old, er, friend.

For more on the trickiness of drug repurposing, see John LaMattina here and here. And the points he raises get to the "as long as it doesn't cost too much" line in the last paragraph. There's opportunity cost involved here, too, of course. When the Broad Institute (or Stanford, or the NIH) screens old pharma candidates for new uses, they're doing what a drug company might do itself, and therefore possibly taking away from work that only they could be doing instead. Now, I think that the Broad (for example) already has a large panel of interesting screens set up, so running the Roche compounds through them couldn't hurt, and might not take that much more time or effort. So why not? But trying to push repurposing too far could end up giving us the worst of both worlds. . .

Comments (14) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

November 28, 2012

Advice For Those Trying High-Throughput Screening

Email This Entry

Posted by Derek

So here's a question that a lot of people around here will have strong opinions on. I've heard from someone in an academic group that's looking into doing some high-throughput screening. As they put it, they don't want to end up as "one of those groups", so they're looking for advice on how to get into this sensibly.

I applaud that; I think it's an excellent idea to look over the potential pitfalls before you hop into an area like this. My first advice would be to think carefully about why you're doing the screening. Are you looking for tool compounds? Do they need to get into cells? Are you thinking of following up with in vivo experiments? Are you (God help you) looking for potential drug candidates? Each of these require somewhat different views of the world.

No matter what, I'd say that you should curate the sorts of structures that you're letting in. Consider the literature on frequent-hitter structures (here's a good starting point, blogged here), and decide how much you want to get hits versus being able to follow up on them. I'd also say to keep in mind the Shoichet work on aggregators (most recently blogged here), especially the lesson that these have to be dealt with assay-by-assay. Compounds that behave normally in one system can be trouble in others - make no assumptions.

But there's a lot more to say about this. What would all of you recommend?

Comments (13) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays

Think Your Drug Is Strange-Looking? Beat This.

Email This Entry

Posted by Derek

We have a late entry in this year's "Least Soluble Molecule - Dosed In Vivo Division" award. Try feeding that into your cLogP program and see what it tells you about its polarity. (This would be a good ChemDraw challenge, too). What we're looking at, I'd say, is a sort of three-dimensional asphalt, decorated around its edges with festive scoops of lard.
The thing is, such structures are perfectly plausible building blocks for various sorts of nanotechnology. It would not, though, have occurred to me to feed any to a rodent. But that's what the authors of this new paper managed to do. The compound shown is wildly fluorescent (as well you might think), and the paper explores its possibilities as an imaging agent. The problem with many - well, most - fluorescent species is photobleaching. That's just the destruction of your glowing molecule by the light used to excite it, and it's a fact of life for almost all the commonly used fluorescent tags. Beat on them enough, and they'll stop emitting light for you.

But this beast is apparently more resistant to photobleaching. (I'll bet it's resistant to a lot of things). Its NMR spectrum is rather unusual - those two protons on the central trypticene show up at 8.26 and 8.91, for example. And in case you're wondering, the M+1 peak in the mass spec comes in at a good solid 2429 mass units, a region of the detector that I'm willing to bet most of us have never explored, or not willingly. The melting point is reported as ">300 C", which is sort of disappointing - I was hoping for something in the four figures.

The paper says, rather drily, that "To direct the biological application of our 3D nanographene, water solubilization is necessary", but that's no small feat. They ended up using Pluronic surfactant, which gave them 100nm particles of the stuff, and they tried these out on both cells and mice. The particles showed very low cytotoxicity (not a foregone conclusion by any means), and were actually internalized to some degree. Subcutaneous injection showed that the compound accumulated in several organs, especially the liver, which is just where you'd expect something like this to pile up. How long it would take to get out of the liver, though, is a good question.

The paper ends with the usual sort of language about using this as a platform for chemotherapy, etc., but I take that as the "insert technologically optimistic conclusion here" macro that a lot of people seem to have loaded into their word processing programs. The main reason this caught my eye is that this is quite possibly the least drug-like molecule I've ever seen actually dosed in an animal. When will we see its like again?

Comments (26) + TrackBacks (0) | Category: Chemical News | Drug Assays

November 27, 2012

How Do Chemist (Think That They) Judge Compounds?

Email This Entry

Posted by Derek

There's an interesting paper out in PLoS One, called "Inside the Mind of a Medicinal Chemist". Now, that's not necessarily a place that everyone wants to go - mine is not exactly a tourist trap, I can tell you - but the authors are a group from Novartis, so they knew what they were getting into. The questions they were trying to answer on this spelunking expedition were:

1) How and to what extent do chemists simplify the problem of identifying promising chemical fragments to move forward in the discovery process? 2) Do different chemists use the same criteria for such decisions? 3) Can chemists accurately report the criteria they use for such decisions?

They took 19 lucky chemists from the Novartis labs and asked them to go through 8 batches of 500 fragments each and select the desirable compounds. For those of you outside the field, that is, unfortunately, a realistic test. We often have to work through lists of this type, for several reasons: "We have X dollars to spend on the screening collection - which compounds should we buy?" "Which of these compounds we already own should still be in the collection, and which should we get rid of?" "Here's the list of screening hits for Enzyme Y: which of these look like useful starting points?" I found myself just yesterday going through about 350 compounds for just this sort of purpose.

They also asked the chemists which of a set of factors they used to make their decisions. These included polarity, size, lipophilicity, rings versus chains, charge, particular functional groups, and so on. Interestingly, once the 19 chemists had made their choices (and reported the criteria they used in doing so), the authors went through the selections using two computational classification algorithms, semi-naïve Bayesian (SNB) and Random Forest (RF). This showed that most of the chemists actually used only one or two categories as important filters, a result that ties in with studies in other fields on how experts in a given subject make decisions. Reducing the complexity of a multifactorial problem is a key step for the human brain to deal with it; how well this reduction is done (trading accuracy for speed) is what can distinguish an expert from someone who's never faced a particular problem before.

But the chemists in this sample didn't all zoom in on the same factors. One chemist showed a strong preference away from the compounds with a higher polar surface area, for example, while another seemed to make size the most important descriptor. The ones using functional groups to pick compounds also showed some individual preferences - one chemist, for example, seemed to downgrade heteroaromatic compounds, unless they also had a carboxylic acid, in which case they moved back up the list. Overall, the most common one-factor preference was ring topology, followed by functional groups and hydrogen bond donors/acceptors.

Comparing structural preferences across the chemists revealed many differences of opinion as well. One of them seemed to like fused six-membered aromatic rings (that would not have been me, had I been in the data set!), while others marked those down. Some tricyclic structures were strongly favored by one chemist, and strongly disfavored by another, which makes me wonder if the authors were tempted to get the two of them together and let them fight it out.

How about the number of compounds passed? Here's the breakdown:

One simple metric of agreement is the fraction of compounds selected by each chemist per batch. The fraction of compounds deemed suitable to carry forward varied widely between chemists, ranging from 7% to 97% (average = 45%), though each chemist was relatively consistent from batch to batch. . .This variance between chemists was not related to their ideal library size (Fig. S7A) nor linearly related to the number of targets a chemist had previously worked on (R2 = 0.05, Fig. S7B). The fraction passed could, however, be explained by each chemist’s reported selection strategy (Fig. S7C). Chemists who reported selecting only the “best” fragments passed a lower fraction of compounds (0.13±0.07) than chemists that reported excluding only the “worst” fragments (0.61±0.34); those who reported intermediate strategies passed an intermediate fraction of compounds (0.39±0.25).

Then comes a key question: how similar were the chemists' picks to each other, or to their own previous selections? A well-known paper from a few years ago suggested that the same chemists, looking at the same list after the passage of time (and more lists!) would pick rather different sets of compounds. Update: see the comments for some interesting inside information on this work.)Here, the authors sprinkled in a couple of hundred compounds that were present in more than one list to test this out. And I'd say that the earlier results were replicated fairly well. Comparing chemists' picks to themselves, the average similarity was only 0.52, which the authors describe, perhaps charitably, as "moderately internally consistent".

But that's a unanimous chorus compared to the consensus between chemists. These had similarities ranging from 0.05 (!) to 0.52, with an average of 0.28. Overall, only 8% of the compounds had the same judgement passed on them by at least 75% of the chemists. And the great majority of those agreements were on bad compounds, as opposed to good ones: only 1% of the compounds were deemed good by at least 75% of the group!

There's one other interesting result to consider: recall that the chemists were asked to state what factors they used in making their decisions. How did those compare to what they actually seemed to find important? (An economist would call this a case of stated preference versus revealed preference). The authors call this an assessment of the chemists' self-awareness, which in my experience, is often a swampy area indeed. And that's what it turned out to be here as well: ". . .every single chemist reported properties that were never identified as important by our SNG or RF classifiers. . .chemist 3 reported that several properties were important, for failed to report that size played any role during selections. Our SNG and RF classifiers both revealed that size, an especially straightforward parameter to assess, was the most important ."

So, what to make of all this? I'd say that it's more proof that we medicinal chemists all come to the lab bench with our own sets of prejudices, based on our own experiences. We're not always aware of them, but they're certainly with us, "sewn into the lining of our lab coats", as Tom Wolfe might have put it. The tricky part is figuring out which of these quirks are actually useful, and how often. . .

Comments (19) + TrackBacks (0) | Category: Drug Assays | Life in the Drug Labs

November 20, 2012

What's A Phenotypic Screen, And What Isn't?

Email This Entry

Posted by Derek

The recent entry here on a phenotypic screen got some discussion going in the comments, and I thought I'd bring that out here for more. Some readers objected to the paper being characterized as a phenotypic screen at all, saying that it was just a cell-based screen. That got me to thinking about how I use the term, and to judge from the comments, there are at least two schools of thought on this.

The first says "Phenotypic" to mean something like "Screening for some desired effect in a living system, independent of any defined target". That's where I come from as well, since I've spent so much of my career doing target-based drug discovery. In a target-based program, you have cell assays, too - but they're downstream of the biochemical/pharmacological assay, and are there to answer two key questions: (1) does hitting the desired target do the right things to the cells, and (2) do the compounds break out into new SAR categories in cells that aren't apparent from their activity against the target? That last part can mean that some of the compounds are cytotoxic (while others aren't), or some of them seem to get into cells a lot better than others, and so on. But they're all subordinated to the original target idea, which drives the whole project.

The other definition of phenotypic screen would be something more like: "Screening simultaneously for a broad range of effects in a living system, independent of any defined target". I would call that, personally, a "high-content" screen (or more precisely, a high-content phenotypic screen, but (as mentioned) opinions vary on this. To the people who think this way, that Broad Institute paper I blogged on was merely a cell assay that looked at the most boring endpoint of all (cell death), and hardly lifted its head beyond that. But to a target-based person, everything that involves throwing compounds onto cells, with no defined target in mind, just to see what happens. . .well, that sure isn't target-based drug discovery, so it must be a phenotypic screen. And death is a phenotype, too, you know.

I like both kinds of screening, just for the record. But they're done for different purposes. High-content screening is a great way to harvest a lot of data and generate a lot of hypotheses, but for drug discovery, it can be a bit too much like a firehose water fountain. A more narrowed-down approach (such as "We want to find some compounds that make these kinds of cells perform Action X") is closer to actionable drug discovery efforts.

At any rate, a reader sent along some good high-content-screening work, and I'll blog about that separately. More comparisons will come up then.

Comments (9) + TrackBacks (0) | Category: Drug Assays

November 15, 2012

A Good Example of Phenotypic Screening

Email This Entry

Posted by Derek

I like to highlight phenotypic screening efforts here sometimes, because there's evidence that they can lead to drugs at a higher-than-usual rate. And who couldn't use some of that? Here's a new example from a team at the Broad Institute.

They're looking at the very popular idea of "cancer stem cells" (CSCs), a population of cells in some tumors that appear to be disproportionately resistant to current therapies (and disproportionately responsible for tumor relapse and regrowth). This screen uses a surrogate breast cell line, with E-cadherin knocked down, which seems to give the dedifferentiated phenotype you'd want to target. That's a bit risky, using an artificial system like that, but as the authors correctly point out, isolating a pure population of the real CSCs is difficult-to-impossible, and they're very poorly behaved in cell culture. So until those problems are solved, you have your choice - work on something that might translate over to the real system, or ditch the screening idea for now entirely. I think the first is worth a shot, as long as its limitations are kept in mind.

This paper does go on to do something very important, though - they use an isogenic cell line as a counterscreen, very close to the target cells. If you find compounds that hit the targets but not these controls, you have a lot more confidence that you're getting at some difference that's tied to the loss of E-cadherin. Using some other cell line as a control leaves too many doors open too wide; you could see "confirmed hits" that are taking advantage of totally irrelevant differences between the cell lines instead.

They ran a library of about 300,000 compounds (the MLSMR collection) past the CSC model cells, and about 3200 had the desired toxic effect on them. At this point, the team removed the compounds that were flagged in PubChem as toxic to normal mammalian cell lines, and also removed compounds that had hit in more than 10% of the assays they'd been through, both of which I'd say are prudent moves. Retesting the remaining 2200 compounds gave a weird result: at the highest concentration (20 micromolar), 97 per cent of them were active. I probably would have gotten nervous at that point, wondering if something had gone haywire with the assay, and I'll bet that a few folks at the Broad felt the same way.

But when used the isogenic cell line, things narrowed down rather quickly. Only 26 compounds showed reasonable potency on the target cells along with at least a 25-fold window for toxicity to the isogenic cells. (Without that screen, then, you'd have been chasing an awful lot of junk). Then they ordered up fresh samples of these, which is another step that believe me, you don't want to neglect. A number of compounds appear to have not been quite what they were supposed to be (not an uncommon problem in a big screening collection; you trust the labels unconditionally at your own peril).

In the end, two acylhydrazone compounds ended up retaining their selectivity after rechecking. So you can see how things narrow down in these situations: 300K to 2K to 26 to 2, and that's not such an unusual progression at all. The team made a series of analogs around the lead chemical matter, and then settled on the acylhydrazone compound shown (ML239) as the best in show. It's not a beauty. There seems to be some rule that more rigorous and unusual a phenotypic screen, the uglier the compounds that emerge from it. I'm only half kidding, or maybe a bit less - there are some issues to think about in there, and that topic is worth a post of its own.
More specifically, the obvious concern in that fulvene-looking pyrrole thingie on the right (I use "thingie" in its strict technical sense here). That's not a happy-looking (that is, particularly stable-looking) group. The acylhydrazine part might raise eyebrows with some people, but Rimonabant (among other compounds) shows that that functional group can be part of a drug. Admittedly, Rimonabant went down with all hands, but it wasn't because of the acylhydrazine. And the trichloroaryl group isn't anyone's favorite, either, but in this context, it's just sort of a dessert topping, in an inverse sense.

But the compound appears to be the real thing, as a pharmacological tool. It was also toxic to another type of breast cancer cell that had had its E-cadherin disrupted, and to a further nonengineered breast cancer cell line. Now comes the question: how does this happen? Gene expression profiling showed a variety of significant changes, with all sorts of cell death and free radical scavenging things altered. By contrast, when they did the same profiling on the isogenic controls, only five genes were altered to any significant extent, and none of those overlapped with the target cells. This is very strong evidence that something specific and important is being targeted here. A closer analysis of all the genes suggests the NF-kappaB system, and within that, perhaps a protein called TRIB3. Further experiments will have to be done to nail that down, but it's a good start. (And yes, in case you were wondering, TRIB3 does, in fact, stand for "tribble-3", and yes, that name did originate with the Drosophila research community, and how did you ever guess?)

So overall, I'd say that this is a very solid example of how phenotypic screening is supposed to work. I recommend it to people who are interested in the topic - and to people who aren't, either, because hey, you never know when it might come in handy. This is how a lot of new biology gets found, through identifying useful chemical matter, and we can never have too much of it.

Comments (23) + TrackBacks (0) | Category: Cancer | Chemical Biology | Drug Assays

November 2, 2012

Sudden Onset of Promiscuity

Email This Entry

Posted by Derek

That title should bring in the hits. But don't get your hopes up! This is medicinal chemistry, after all.

"Can't you just put the group in your molecule that does such-and-such?" Medicinal chemists sometimes hear variations of that question from people outside of chemistry - hopeful sorts who believe that we might have some effective and instantly applicable techniques for fixing selectivity, brain penetration, toxicity, and all those other properties we're always trying to align.

Mostly, though, we just have general guidelines - not so big, not so greasy (maybe not so polar, either, depending on what you're after), and avoid a few of the weirder functional groups. After that, it's art and science and hard work. A recent J. Med. Chem. paper illustrates just that point - the authors are looking at the phenomenon of molecular promiscuity. That shows up sometimes when one compound is reasonably selective, but a seemingly closely related one hits several other targets. Is there any way to predict this sort of thing?

"Probably not", is the answer. The authors looked at a range of matched molecular pairs (MMPs), structures that were mostly identical but varied only in one region. Their data set is list of compounds in this paper from the Broad Institute, which I blogged about here. There are over 15,000 compounds from three sources - vendors, natural product collections, and Schreiber-style diversity-oriented synthesis. The MMPs are things like chloro-for-methoxy on an aryl ring, or thiophene-for-pyridyl with other substituents the same. That is, they're just the sort of combinations that show up when medicinal chemists work out a series of analogs.

The Broad data set yielded 30954 matched pairs, involving over 8000 compounds and over seven thousand different transformations. Comparing these compounds and their reported selectivity over 100 different targets (also in the original paper), showed that most of these behaved "normally" - over half of them were active against the same targets that their partners were active against. But at the other end of the scale, 829 compounds showed different activity over at least ten targets, and 126 of those compounds different in activity by fifty targets or more. 33 of them differed by over ninety targets! So there really are some sudden changes out there waiting to be tripped over; they're not frequent, but they're dramatic.

How about correlations between these "promiscuity cliff" compounds and physical properties, such as molecular weight, logP, donor/acceptor count, and so on? I'd have guessed that a change to higher logP would have accompanied this sort of thing over a broad data set, but the matched pairs don't really show that (nor a shift in molecular weight). On the other hand, most of the highly promiscuous compounds are in the high cLogP range, which is reassuring from the standpoint of Received Med-Chem Wisdom. There are still plenty of selective high-logP compounds, but the ones that hit dozens of targets are almost invariably logP > 6.

Structurally, though, no particular substructure (or transformation of substructures) was found to be associated with sudden onset of promiscuity, so to this approximation, there's no actionable "avoid sticking this thing on" rule to be drawn. (Note that this does not, to me at least, say that there are no such things are frequent-hitting structures - we're talking about changes within some larger structure, not the hits you'd get when screening 500 small rhodanine phenols or the like). In fact, I don't think the Broad data set even included many functional groups of that sort to start with.

The take-home:

On the basis of the data available to us, it is not possible to conclude with certainty to what extent highly promiscuous compounds engage in specific and/or nonspecific interactions with targets. It is of course unlikely that a compound might form specific interactions with 90 or more diverse targets, even if the interactions were clearly detectable under the given experimental conditions. . .

. . .it has remained largely unclear from a medicinal chemistry perspective thus far whether certain molecular frameworks carry an intrinsic likelihood of promiscuity and/or might have frequent hitter character. After all, promiscuity is determined for compounds, not their frameworks. Importantly, the findings presented herein do not promote a framework-centric view of promiscuity. Thus, for the evaluation and prioritization of compound series for medicinal chemistry, frameworks should not primarily be considered as an intrinsic source of promiscuity and potential lack of compound specificity. Rather, we demonstrate that small chemical modifications can trigger large-magnitude promiscuity effects. Importantly, these effects depend on the specific structural environment in which these modifications occur. On the basis of our analysis, substitutions that induce promiscuity in any structural environment were not identified. Thus, in medicinal chemistry, it is important to evaluate promiscuity for individual compounds in series that are preferred from an SAR perspective; observed specificity of certain analogs within a series does not guarantee that others are not highly promiscuous."

Point taken. I continue to think, though, that some structures should trigger those evaluations with more urgency than others, although it's important never to take anything for granted with molecules you really care about.

Comments (10) + TrackBacks (0) | Category: Chemical News | Drug Assays | Natural Products | Toxicology

October 12, 2012

Controversial Scaffold Is Right

Email This Entry

Posted by Derek

Let's talk tool compounds. This topic has come up around here before, generally when some paper gets published from an academic group featuring a hideous molecule. Today, alas, is no exception. Feast your eyes on this one, an inhibitor of tyrosyl-DNA phosphodiesterase I.
Now, on one level, I'm sympathetic. That's an unusual enzyme, and there really aren't any decent inhibitors known for it. It's pretty hard to work out the function of an enzyme without a good inhibitor to watch the effects in living cells, and in that sense, it's good to have found something. But here comes the other hand: your compound for such a purpose needs to be a good one, or the studies you run with it risk being meaningless. "Good" is a relative term, I realize, depending on how much you're expecting the compound to do for you: enzyme assay only? Cells? Mice? But at a minimum, it should be selective for the target you're trying to figure out.

And that's where my eyebrows go up with this little beast. It's not quite a rhodanine (switch the carbonyl and thiocarbonyl if you want that), but it's not a particularly pretty heterocycle, either. This paper, which I wrote about here, looked at the promiscuity of several related heterocycles, but not this one in particular. Any time I see a thioamide group, I get worried. This very system, in fact, shows up in the PAINS paper (open access copy), which I blogged about here, warns against this whole class (see page S45 of the supplementary material).

It also warns, with good reason, against the alkylidene branching off of such rings. I know that there are drugs with such features (epalrestat, Sutent), but your chances of such compounds being real and going all the way are surely lower. Finally, we come to the triphenol. Polyphenol compounds are notorious in medicinal chemistry. They're reactive, they're unstable, they show up in all kinds of assays, and their SAR almost never makes any sense. So this compound has lots going for it.

The authors realize this, and checked the compounds against bovine serum albumin as a way to assay nonspecific protein binding. They also did some work with whole-cell extracts, and continue to feel that these "controversial scaffolds" (their words) can still be useful. (And to be fair, they're also looking at replacing the phenolic section with less nasty polar groups). But while their hearts (and their heads) are in the right place, I still worry very much about these compounds. I'd be quite interested in seeing them run across a broad panel of assays, just to see how promiscuous they really are, and I would be very careful about trusting cellular data (or anything close to it) until that's done.

Comments (24) + TrackBacks (0) | Category: Drug Assays

September 26, 2012

Free the Labels

Email This Entry

Posted by Derek

When you talking assays, "label-free" is a magic phrase. The more thingies you have to stick onto your molecules or targets to see them, the less confidence you'll have that you're actually looking at the system the way you really wanted to see it: as if you weren't looking at it at all. And while we're not quite quantum mechanics, the observer effect is very real in molecular and cell biology - too many interesting techniques perturb the system in the process of reading out.

And there are no perfect label-free assays, otherwise we'd all be using them. In vitro, NMR can tell you an awful lot, but it can require an awful lot of work if you want to correlate structural information with binding events. And mass spec is getting ridiculously sensitive, and can be used to detect compound binding. But even when that works, it doesn't give you any structure (or much spatial resolution after a certain point, if that's what you're looking for - say, in cells). SPR is a great technique for getting kinetic information right out of the primary assay (instant off-rates!) But it's not quite label-free, because you have to immobilize something to a chip to make it work. Thermal shift is an interesting assay, too - but it uses up a fair amount of protein, and some proteins are more sensitive to it than others. No structural information there, either.

There are a couple of techniques that I don't have much experience with that sound intriguing. Capillary electrophoresis for binding is one - you look at mobility changes with your protein when something is bound to it, as you'd imagine. It's supposed to be pretty sensitive. And BLI (bio-layer interferometry) reminds me a bit of SPR, in that it uses an immobilized protein. I'm not sure what the advantages/disadvantages of that one are, but I see it turn up in the literature.

The ideal assay? If you could do NMR, with the sensitivity to detect very small amounts of a compound, with spatial resolution well below subcellular. You'd get binding, localization, and structure all in one shot. That's probably not even possible, but I'd love to be wrong about that.

Comments (15) + TrackBacks (0) | Category: Drug Assays

August 22, 2012

Watch that Little Letter "c"

Email This Entry

Posted by Derek

Hang around a bunch of medicinal chemists (no, really, it's more fun than you'd think) and you're bound to hear discussion of cLogP. For the chemists in the crowd, I should warn you that I'm about to say nasty things about it.

For the nonchemists in the crowd, logP is a measure of how greasy (or how polar) a compound is. It's based on a partition experiment: shake up a measured amount of a compound with defined volumes of water and n-octanol, a rather greasy solvent which I've never seen referred to in any other experimental technique. Then measure how much of the compound ends up in each layer, and take the log of the octanol/water ratio. So if a thousand times as much compound goes into the octanol as goes into the water (which for drug substances is quite common, in fact, pretty good), then the logP is 3. The reason we care about this is that really greasy compounds (and one can go up to 4, 5, 6, and possibly beyond), have problems. They tend to dissolve poorly in the gut, have problems crossing membranes in living systems, get metabolized extensively in the liver, and stick to a lot of proteins that you'd rather they didn't stick to. Fewer high-logP compounds are capable of making it as drugs.

So far, so good. But there are complications. For one thing, that description above ignores the pH of the water solution, and for charged compounds that's a big factor. logD is the term for the distribution of all species (ionized or not), and logD at pH 7.4 (physiological) is a valuable measurement if you've got the possibility of a charged species (and plenty of drug molecules do, thanks to basic amines, carboxylic acids, etc.) But there are bigger problems.

You'll notice that the experiment outlined in the second paragraph could fairly be described as tedious. In fact, I have never seen it performed. Not once, and I'll bet that the majority of medicinal chemists never have, either. And it's not like it's just being done out of my sight; there's no roomful of automated octanol/water extraction machines clanking away in the basement. I should note that there are other higher-throughput experimental techniques (such as HPLC retention times) that also correlate with logP and have been used to generate real numbers, but even those don't account for the great majority of the numbers that we talk about all the time. So how do we manage to do that?

It has to do with a sleight of hand I've performed while writing the above sections, which some of you have probably already noticed. Most of the time, when we talk about logP values in early drug discovery, we're talking about cLogp. That "c" stands for calculated. There are several programs that estimate logP based on known values for different rings and functional groups, and with different algorithms for combining and interpolating them. In my experience, almost all logP numbers that get thrown around are from these tools; no octanol is involved.

And sometimes that worries me a bit. Not all of these programs will tell you how solid those estimates are. And even if they will, not all chemists will bother to check. If your structure is quite close to something that's been measured, then fine, the estimate is bound to be pretty good. But what if you feed in a heterocycle that's not in the lookup table? The program will spit out a number, that's what. But it may not be a very good number, even if it goes out to two decimal places. I can't even remember when I might have last seen a cLogP value with a range on it, or any other suggestion that it might be a bit fuzzy.

There are more subtle problems, too - I've seen some oddities with substitutions on saturated heterocyclic rings (morpholine, etc.) that didn't quite seem to make sense. Many chemists get these numbers, look at them quizzically, and say "Hmm, I didn't know that those things sorted out like that. Live and learn!" In other words, they take the calculated values as reality. I've even had people defend these numbers by explaining to me patiently that these are, after all, calculated logP values, and the calculated log P values rank-order like so, and what exactly is my problem? And while it's hard to argue with that, we are not putting our compounds into the simulated stomachs of rationalized rodents. Real-world decisions can be made based on numbers that do not come from the real world.

Comments (39) + TrackBacks (0) | Category: Drug Assays | In Silico | Life in the Drug Labs

August 21, 2012

Four Billion Compounds At a Time

Email This Entry

Posted by Derek

This paper from GlaxoSmithKline uses a technology that I find very interesting, but it's one that I still have many questions about. It's applied in this case to ADAMTS-5, a metalloprotease enzyme, but I'm not going to talk about the target at all, but rather, the techniques used to screen it. The paper's acronym for it is ELT, Encoded Library Technology, but that "E" could just as well stand for "Enormous".

That's because they screened a four billion member library against the enzyme. That is many times the number of discrete chemical species that have been described in the entire scientific literature, in case you're wondering. This is done, as some of you may have already guessed, by DNA encoding. There's really no other way; no one has a multibillion-member library formatted in screening plates and ready to go.

So what's DNA encoding? What you do, roughly, is produce a combinatorial diversity set of compounds while they're attached to a length of DNA. Each synthetic step along the way is marked by adding another DNA sequence to the tag, so (in theory) every compound in the collection ends up with a unique oligonucleotide "bar code" attached to it. You screen this collection, narrow down on which compound (or compounds) are hits, and then use PCR and sequencing to figure out what their structures must have been.

As you can see, the only way this can work is through the magic of molecular biology. There are so many enzymatic methods for manipulating DNA sequences, and they work so well compared with standard organic chemistry, that ridiculously small amounts of DNA can be detected, amplified, sequenced, and worked with. And that's what lets you make a billion member library; none of the components can be present in very much quantity (!)
This particular library comes off of a 1,3,5-triazine, which is not exactly the most cutting-edge chemical scaffold out there (I well recall people making collections of such things back in about 1992). But here's where one of the big questions comes up: what if you have four billion of the things? What sort of low hit rate can you not overcome by that kind of brute force? My thought whenever I see these gigantic encoded libraries is that the whole field might as well be called "Return of Combichem: This Time It Works", and that's what I'd like to know: does it?

There are other questions. I've always wondered about the behavior of these tagged molecules in screening assays, since I picture the organic molecule itself as about the size of a window air conditioner poking out from the side of a two-story house of DNA. It seems strange to me that these beasts can interact with protein targets in ways that can be reliably reproduced once the huge wad of DNA is no longer present, but I've been assured by several people that this is indeed the case.

In this example, two particular lineages of compounds stood out as hits, which makes you much happier than a collection of random singletons. When the team prepared a selection of these as off-DNA "real organic compounds", many of them were indeed nanomolar hits, although a few dropped out. Interestingly, none of the compounds had the sorts of zinc-binding groups that you'd expect against the metalloprotease target. The rest of the paper is a more traditional SAR exploration of these, leading to what one has to infer are more tool/target validation compounds rather than drug candidates per se.

I know that GSK has been doing this sort of thing for a while, and from the looks of it, this work itself was done a while ago. For one thing, it's in J. Med. Chem., which is not where anything hot off the lab bench appears. For another, several of the authors of the paper appear with "Present Address" footnotes, so there has been time for a number of people on this project to have moved on completely. And that brings up the last set of questions, for now: has this been a worthwhile effort for GSK? Are they still doing it? Are we just seeing the tip of a large and interesting iceberg, or are we seeing the best that they've been able to do? That's the drug industry for you; you never know how many cards have been turned over, or why.

Comments (24) + TrackBacks (0) | Category: Chemical Biology | Chemical News | Drug Assays | Drug Industry History

August 3, 2012

Finding Fast Fruit Fly Feasibility

Email This Entry

Posted by Derek

I'm an unabashed fan of phenotypic screening. (For those outside the field, that means screening for compounds by looking for their effects on living systems, rather than starting with a molecular target and working your way up). Done right, I don't think that there's a better platform for breakthrough drug discovery, mainly because there's so much we don't know about what really goes on in cells and in whole organisms.

Doing it right isn't easy, though, nor will you necessarily find anything even if you do. But there's a recent paper in Nature that is, I think, a model of the sort of thing that we should all be thinking about. A collaboration between the Shokat group at UCSF and the Cagan group at Mt. Sinai, this project is deliberately looking for one at the trickiest aspects of drug discovery: polypharmacology. "One target, one drug" is all very well, but what if your drugs hit more than one target (as they generally do?) Or what if your patients will only be served by hitting more than one target (as many diseases, especially cancer, call for)? The complexities get out of control very quickly, and model systems would be very helpful indeed.

This work goes all the way back to fruit flies, good ol' Drosophila, and the authors picked a well-characterized cancer pathway: multiple endocrine neoplasia type 2 (MEN2). This is known to be driven by gain-of-function mutations in the Ret pathway, and patients with such mutations show a greatly increased rate of endocrine tumors (thyroid, especially). Ret is a receptor tyrosine kinase, and the receptor is one that recognizes the GDNF family of signaling peptides. As oncology pathways go, this one is fairly well worked out, not that it's led to any selective Ret inhibitor drugs so far (although many have tried and are trying).

Using this Ret-driven fly model, the teams ran a wide variety of kinase inhibitor molecules past the insects, looking for their effects, while at the same time profiling the compounds across a long list of kinase enzymes. This gives you a chance to do something that you don't often get a chance to do: match one kind of fingerprint to another kind. And what they found was that you needed "balanced polypharmacology" to get optimal phenotypic effects. The compounds that inhibited the Drosophila equivalents of Ret, Raf, Src and S6K all at the same time made the flies survive the longest. That's quite a blunderbuss list. But some very similar compounds weren't as good, and that turned out to be due to the activity on Tor. Working these combinations out was not trivial - it took a lot of different strains of flies with different levels of kinase activity, and a lot of different compounds with varying profiles.

Now, these kinases cover an awful lot of ground, as you'll know if you've worked in the field, or if you just click on those links and look at some of the pathway diagrams. There is, I think it's fair to say, no way that anyone could have identified these particular combinations with certainly without running the experiment in a real system; there are just too many branching, intersecting, ramifications to get a clear picture of what would happen. Thus, phenotypic screening: let the real system tell you.

So, you may be thinking, fruit flies. Great. Does that tell us anything real? In this case, it looks like it does. The compound profiles that were seen in the model system translated to human cell lines, and to mouse xenograft models. And while neither of those is a perfect indicator (far from it), they're about the best we have, and many are the compounds that have gone into human trials with just such data.

I look forward to more applications of this technique, to see how far it can be pushed. Ret looks like a well-chosen test case - what happens when you go on to even trickier ones? It won't be easy, but being able to unravel some of the polypharmacology when you're still back at the fruit-fly stage will be worth the effort.

Comments (9) + TrackBacks (0) | Category: Cancer | Drug Assays

August 2, 2012

Public Domain Databases in Medicinal Chemistry

Email This Entry

Posted by Derek

Here's a useful overview of the public-domain medicinal chemistry databases out there. It covers the big three databases in detail:

BindingDB (quantitative binding data to protein targets).

ChEMBL (wide range of med-chem data, overlaps a bit with PubChem).

PubChem (data from NIH Roadmap screen and many others).

And these others:
Binding MOAD (literature-annotated PDB data).

ChemSpider (26 million compounds from hundreds of data sources).

DrugBank (data on 6700 known drugs).

GRAC and IUPHAR-DB (data on GPCRs, ion channels, and nuclear receptors, and ligands for all of these).

PDBbind (more annotated PDB data).

PDSP Ki (data from UNC's psychoactive drug screening program)

SuperTarget (target-compound interaction database).

Therapeutic Targets Database(database of known and possible drug targets).

ZINC (21 million commercially available compounds, organized by class, downloadable in various formats).

There is the irony of a detail article on public-domain databases appearing behind the ACS paywall, but the literature is full of such moments as that. . .

Comments (11) + TrackBacks (0) | Category: Biological News | Chemical News | Drug Assays

July 27, 2012

Antipsychotic Drugs Against Cancer?

Email This Entry

Posted by Derek

One of the hazards of medicinal chemistry - or should I say, one of the hazards of long experience in medicinal chemistry - is that you start to think that you know more than you do. Specifically, after a few years and a few projects, you've seen plenty of different compounds and their activities (or lack thereof). Human brains categorize things and seek patterns, so it's only natural that you develop a mental map of the chemical space you've encountered. Problem is, any such map has to be incomplete, grievously incomplete, and if you start making too many decisions based on it (rather than on actual data), you can miss out on some very useful things.

Here's a case in point: an assay against cancer stem cells, which have been a hot research area for some time now. It may well be that some classes of tumor are initiated and then driven by such cells, in which case killing them off or inactivating them would be a very good thing indeed. This was an interesting assay, because it included control stem cells to try to differentiate between compounds that would have an effect on the neoplasm-derived cells while leaving the normal ones alone.

And what did they find? Thioridiazine is what - an old-fashioned phenothiazine antipsychotic drug. For reasons unknown, it's active against these cancer stem cells. When the authors did follow-up screening, two other compounds of this class also showed up active: fluphenazine and prochlorperazine, so I'd certainly say that this is real.

And it appears that it might actually be the compounds' activity against dopamine receptors that drives this assay. The authors found that there's a range of dopamine receptor expression in such cells, and that this correlates with the activity of the phenothiazine compounds. That's quite interesting, but it complicates life quite a bit for running assays:

Our observations of differential DR expression between normal and neoplastic patient samples strongly suggest human CSCs are heterogeneous and drug targeting should be based on molecular pathways instead of surrogate phenotypic markers.

Working out molecular pathways is hard; a lot more progress might be made at this stage of the game by running phenotypic assays - but not if they're against a heterogeneous cell population. That way lies madness.

Interesting, the phenothiazines had been reported to show some anti-cancer effects, and schizophrenic patients receiving such drugs had been reported to show lower incidences of some forms of cancer. These latest observations might well be the link between all these things, and seem to represent the only tractable small-molecule approach (so far) targeting human cancer stem cells.

But you have to cast your net wide to find such things. Dopamine receptors aren't the most obvious thing to suspect here, and ancient antipsychotics aren't the most obvious chemical matter to screen. Drop your preconceptions at the door, is my advice.

Comments (22) + TrackBacks (0) | Category: Cancer | Drug Assays | The Central Nervous System

July 23, 2012

Science Fiction Gets the Upper Hand

Email This Entry

Posted by Derek

I wrote here about the Cronin lab at Glasgow and their work on using 3-D printing technology to make small chemical reactors. Now there's an article on this research in the Observer that's getting some press attention (several people have e-mailed it to me). Unfortunately, the headline gets across the tone of the whole piece: "The 'Chemputer' That Could Print Out Any Drug".

To be fair, this was a team effort. As the reporter notes, Prof. Cronin "has a gift for extrapolation", and that seems to be a fair statement. I think that such gifts have to be watched carefully in the presence of journalists, though. The whole story is a mixture of wonderful-things-coming-soon! and still-early-days-lots-of-work-to-be-done, and these two ingredients keep trying to separate and form different layers:

So far Cronin's lab has been creating quite straightforward reaction chambers, and simple three-step sequences of reactions to "print" inorganic molecules. The next stage, also successfully demonstrated, and where things start to get interesting, is the ability to "print" catalysts into the walls of the reactionware. Much further down the line – Cronin has a gift for extrapolation – he envisages far more complex reactor environments, which would enable chemistry to be done "in the presence of a liver cell that has cancer, or a newly identified superbug", with all the implications that might have for drug research.

In the shorter term, his team is looking at ways in which relatively simple drugs – ibuprofen is the example they are using – might be successfully produced in their 3D printer or portable "chemputer". If that principle can be established, then the possibilities suddenly seem endless. "Imagine your printer like a refrigerator that is full of all the ingredients you might require to make any dish in Jamie Oliver's new book," Cronin says. "Jamie has made all those recipes in his own kitchen and validated them. If you apply that idea to making drugs, you have all your ingredients and you follow a recipe that a drug company gives you. They will have validated that recipe in their lab. And when you have downloaded it and enabled the printer to read the software it will work. The value is in the recipe, not in the manufacture. It is an app, essentially."

What would this mean? Well for a start it would potentially democratise complex chemistry, and allow drugs not only to be distributed anywhere in the world but created at the point of need. It could reverse the trend, Cronin suggests, for ineffective counterfeit drugs (often anti-malarials or anti-retrovirals) that have flooded some markets in the developing world, by offering a cheap medicine-making platform that could validate a drug made according to the pharmaceutical company's "software". Crucially, it would potentially enable a greater range of drugs to be produced. "There are loads of drugs out there that aren't available," Cronin says, "because the population that needs them is not big enough, or not rich enough. This model changes that economy of scale; it could makes any drug cost effective."

Not surprisingly Cronin is excited by these prospects, though he continually adds the caveat that they are still essentially at the "science fiction" stage of this process. . .

Unfortunately, "science fiction" isn't necessarily a "stage" in some implied process. Sometimes things just stay fictional. Cronin's ideas are not crazy, but there are a lot of details between here and there, and if you don't know much organic chemistry (as many of the readers of the original article won't), then you probably won't realize how much work remains to be done. Here's just a bit; many readers of this blog will have thought of these and more:

First, you have to get a process worked out for each of these compounds, which will require quite a bit of experimentation. Not all reagents and solvents are compatible with the silicone material that these microreactors are being fabricated from. Then you have to ask yourself, where do the reagents and raw materials come in? Printer cartridges full of acetic anhydride and the like? Is it better to have these shipped around and stored than it is to have the end product? In what form is the final drug produced? Does it drip out the end of the microreactor (and in what solvent?), or is a a smear on some solid matrix? Is it suitable for dosing? How do you know how much you've produced? How do you check purity from batch to batch - in other words, is there any way of knowing if something has gone wrong? What about medicines that need to be micronized, coated, or treated in the many other ways that pills are prepared for human use?

And those are just the practical considerations - some of them. Backing up to some of Prof. Cronin's earlier statements, what exactly are those "loads of drugs out there that aren't available because the population that needs them is not big enough, or not rich enough"? Those would be ones that haven't been discovered yet, because it's not like we in the industry have the shelves lined with compounds that work that we aren't doing anything with for some reason. (Lots of people seem to think that, though). Even if these microreactors turn out to be a good way to make compounds, though, making compounds has not been the rate-limiting step in discovering new drugs. I'd say that biological understanding is a bigger one, or (short of that), just having truly useful assays to find the compounds you really want.

Cronin has some speculations on that, too - he wonders about the possibility of having these microreactors in some sort of cellular or tissue environment, thus speeding up the whole synthesis/assay loop. That would be a good thing, but the number of steps that have to be filled in to get that to work is even larger than for the drug-manufacture-on-site idea. I think it's well worth working on - but I also think it's well worth keeping out of the newspapers just yet, too, until there's something more to report.

Comments (29) + TrackBacks (0) | Category: Academia (vs. Industry) | Chemical News | Drug Assays

June 25, 2012

A Kinase Inhibitor Learns Something New

Email This Entry

Posted by Derek

Here's another reminder that we don't know what a lot of existing drugs are doing on the side. This paper reports that the kinase inhibitor Nexavar (sorafenib) is actually a pretty good ligand at 5-HT (serotinergic) receptors, which is not something that you'd have guessed at all.

The authors worked up a binding model for the 5-HT2a receptor and ran through lists of known drugs. Sorafenib was flagged, and was (experimentally) a 2 micromolar antagonist. As it turns out, though, it's an even strong ligand for 5-HT2b (57 nM!) and 5-HT2c (417 nM), with weaker activity on a few other subtypes. This makes a person wonder about the other amine GPCR receptors, since there's often some cross-reactivity with small molecule ligands. (Those, though, often have good basic tertiary amines in them, carrying a positive charge under in vivo conditions. Sorafenib lacks any such thing, so it'll be interesting to see the results of further testing). It's also worth wondering if these serotinergic activities help or hurt the drug in oncology indications. In case you're wondering, the compound does get into the brain, although it's significantly effluxed by the BCRP transporter.

What I also find interesting is that this doesn't seem to have been picked up by some of the recent reports on attempts to predict and data-mine potential side effects. We still have a lot to learn, in case anyone had any doubts.

Comments (18) + TrackBacks (0) | Category: Cancer | Drug Assays | The Central Nervous System | Toxicology

June 19, 2012

Watch Your Cell Assays

Email This Entry

Posted by Derek

I've written here before about the (now) well-known problem of compound aggregation in screening assays. You can get false positives when a compound itself forms a colloidal mass that pulls the target protein into it. The readout looks as if the protein has been inactivated by the small molecule itself, just the way you were hoping for, but if you add a bit of detergent the activity mysteriously goes away.

The Shoichet lab has a paper out that warns people to look out for this in cellular assays as well. This time you'll get false negatives - the colloidal aggregates don't act right compared to the free molecules, as you could well imagine. Update: I see that Wavefunction has covered this same paper! Reformulating the assay restores the activity, but the trick is knowing that there was a problem to start with. Something to keep in mind when your cell assay numbers are wonky (as if there weren't enough reasons for that to happen already).

Comments (21) + TrackBacks (0) | Category: Drug Assays

June 1, 2012

Return of the Rhodanome

Email This Entry

Posted by Derek

I do hate to bring up rhodanines again, but I'm not the one who keeps making the things. This paper from ACS Medicinal Chemistry Letters turns out dozens of the things as potential inhibitors of the cellular protein dynamin, in what a colleague of mine referred to as a "nice exploration of the rhodanome".

He did not say it with a straight face. But this paper does: "The rhodanine core is a privileged scaffold in medicinal chemistry and one that has found promise among many therapeutic applications." Well, that's one way to look at it. Another viewpoint is that rhodanines are "polluting the scientific literature" and that they should "be considered very critically" no matter what activity they show in your assay.

The usual answer to this is that these aren't drugs, they're tool compounds. But I don't think that these structures even make safe tools; they have the potential to do too many other things in cell assays. But if people are going to go ahead and use them, I wish that they'd at least make a nod in that direction, instead of mentioning, in passing, how great the whole class is. And yes, I know that they cite two papers to that effect, but one of those two mainly just references the other one when it comes to rhodanines. My viewpoint is more like this paper's:

Academic drug discovery is being accompanied by a plethora of publications that report screening hits as good starting points for drug discovery or as useful tool compounds, whereas in many cases this is not so. These compounds may be protein-reactive but can also interfere in bioassays via a number of other means, and it can be very hard to prove early on that they represent false starts. . .

And I endorse this view as well:

. . .Barriers to adoption of best practices for some academic drug-discovery researchers include knowledge gaps and infrastructure deficiencies, but they also arise from fundamental differences in how academic research is structured and how success is measured. Academic drug discovery should not seek to become identical to commercial pharmaceutical research, but we can do a better job of assessing and communicating the true potential of the drug leads we publish, thereby reducing the wastage of resources on nonviable compounds.

Comments (19) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays

May 23, 2012

Another Vote Against Rhodanines

Email This Entry

Posted by Derek

For those of you who'd had to explain to colleagues (in biology or chemistry) why you're not enthusiastic about the rhodanine compounds that came out of your high-throughput screening effort, there's now another paper to point them to.

The biological activity of compounds possessing a rhodanine moiety should be considered very critically despite the convincing data obtained in biological assays. In addition to the lack of selectivity, unusual structure–activity relationship profiles and safety and specificity problems mean that rhodanines are generally not optimizable.

That's well put, I think, although this has been a subject of debate. I would apply the same language to the other "PAINS" mentioned in the Baell and Holloway paper, which brought together a number of motifs that have set off alarm bells over the years. These structures are guilty until proven innocent. If you have a high-value target and feel that it's worth the time and trouble to prove them so, that may well be the right decision. But if you have something else to advance, you're better off doing so. As I've said here before, ars longa, pecunia brevis.

Comments (3) + TrackBacks (0) | Category: Drug Assays | Drug Development

May 22, 2012

The NIH's Drug Repurposing Initiative: Will It Be a Waste?

Email This Entry

Posted by Derek

The NIH's attempt to repurpose shelved development compounds and other older drugs is underway:

The National Institutes of Health (NIH) today announced a new plan for boosting drug development: It has reached a deal with three major pharmaceutical companies to share abandoned experimental drugs with academic researchers so they can look for new uses. NIH is putting up $20 million for grants to study the drugs.

"The goal is simple: to see whether we can teach old drugs new tricks," said Health and Human Services Secretary Kathleen Sebelius at a press conference today that included officials from Pfizer, AstraZeneca, and Eli Lilly. These companies will give researchers access to two dozen compounds that passed through safety studies but didn't make it beyond mid-stage clinical trials. They shelved the drugs either because they didn't work well enough on the disease for which they were developed or because a business decision sidelined them.

There are plenty more where those came from, and I certainly wish people luck finding uses for them. But I've no idea what the chances for success might be. On the one hand, having a compound that's passed all the preclinical stages of development and has then been into humans is no small thing. On that ever-present other hand, though, randomly throwing these compounds against unrelated diseases is unlikely to give you anything (there aren't enough of them to do that). My best guess is that they have a shot in closely related disease fields - but then again, testing widely might show us that there are diseases that we didn't realized were related to each other.

John LaMattina is skeptical:

Well, the NIH has recently expanded the remit of NCATS. NCATS will now be testing drugs that have been shelved by the pharmaceutical industry for other potential uses. The motivation for this is simple. They believe that these once promising but failed compounds could have other uses that the inventor companies haven’t yet identified. I’d like to reiterate the view of Dr. Vagelos – it’s fairy time again.

My views on this sort of initiative, which goes by a variety of names – “drug repurposing,” “drug repositioning,” “reusable drugs” – have been previously discussed in my blog. I do hope that people can have success in this type of work. But I believe successes are going to be rare.

The big question is, rare enough to count the money and time as wasted, or not? I guess we'll find out. Overall, I'd rather start with a compound that I know does what I want it to do, and then try to turn it into a drug (phenotypic screening). Starting with a compound that you know is a drug, but doesn't necessarily do what you want it to, is going to be tricky.

Comments (33) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development | Drug Industry History

May 21, 2012

A New Way to Kill Amoebas, From An Old Drug

Email This Entry

Posted by Derek

Here's a good example of phenotypic screening coming through with something interesting and worthwhile: they screened against Entamoeba histolytica, the protozooan that causes amoebic dysentery and kills tens of thousands of people every year. (Press coverage here).

It wasn't easy. The organism is an anaerobe, which is a bad fit for most robotic equipment, and engineering a decent readout for the assay wasn't straightforward, either. They did have a good positive control, though - the nitroimidazole drug metronidazole, which is the only agent approved currently against the parasite (and to which it's becoming resistant). A screen of nearly a thousand known drugs and bioactive compounds showed eleven hits, of which one (auranofin) was much more active than metronidazole itself.

Auranofin's an old arthritis drug. It's a believable result, because the compound has also been shown to have activity against trypanosomes, Leishmania parasites, and Plasmodium malaria parasites. This broad-spectrum activity makes some sense when you realize that the drug's main function is to serve as a delivery vehicle for elemental gold, whose activity in arthritis is well-documented but largely unexplained. (That activity is also the basis for persistent theories that arthritis may have an infectious-disease component).

The target in this case may well be arsenite-inducible RNA-associated protein (AIRAP), which was strongly induced by drug treatment. The paper notes that arsenite and auranofin are both known inhibitors of thioredoxin reductase, which strongly suggests that this is the mechanistic target here. The organism's anaerobic lifestyle fits in with that; this enzyme would presumably be its main (perhaps only) path for scavenging reactive oxygen species. It has a number of important cysteine residues, which are very plausible candidates for binding to a metal like gold. And sure enough, auranofin (and two analogs) are potent inhibitors of purified form of the amoeba enzyme.

The paper takes the story all the way to animal models, where auranofin completely outperforms metronidazole. The FDA has now given it orphan-drug status for amebiasis, and the way appears clear for a completely new therapeutic option in this disease. Congratulations to all involved; this is excellent work.

Comments (10) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development | Infectious Diseases

May 16, 2012

Antidepressant Drugs and Cell Membranes

Email This Entry

Posted by Derek

How much do we really know about what small drug molecules do when they get into cells? Everyone involved in this sort of research wonders about this question, especially when it comes to toxicology. There's a new paper out in PLoS One that will cause you to think even harder.

The researchers (from Princeton) looked at the effects of the antidepressant sertraline, a serotonin reuptake inhibitor. They did a careful study in yeast cells on its effects, and that may have some of you raising your eyebrows already. That's because yeast doesn't even have a serotonin transporter. In a perfect pharmacological world, sertraline would do nothing at all in this system.

We don't live in that world. The group found that the drug does enter yeast cells, mostly by diffusion, with a bit of acceleration due to proton motive force and some reverse transport by efflux pumps. (This is worth considering in light of those discussions we were having here the other day about transport into cells). At equilibrium, most (85 to 90%) of the sertaline that makes it into a yeast cell is stuck to various membranes, mostly ones involved in vesicle formation, either through electrostatic forces or buried in the lipid bilayer. It's not setting off any receptors - there aren't any - so what happens when it's just hanging around in there?

More than you'd think, apparently. There's enough drug in there to make some of the membranes curve abnormally, which triggers a local autophagic response. (The paper has electron micrographs of funny-looking Golgi membranes and other organelles). This apparently accounts for the odd fact, noticed several years ago, that some serotonin reuptake inhibitors have antifungal activity. This probably applies to the whole class of cationic amphiphilic/amphipathic drug structures.

The big question is what happens in mammalian cells at normal doses of such compounds. These may well not be enough to cause membrane trouble, but there's already evidence to the contrary. A second big question is: does this effect account for some of the actual neurological effects of these drugs? And a third one is, how many other compounds are doing something similar? The more you look, the more you find. . .

Comments (25) + TrackBacks (0) | Category: Drug Assays | Pharmacokinetics | The Central Nervous System | Toxicology

April 27, 2012

How Do Drugs Get Into Cells? A Vicious Debate.

Email This Entry

Posted by Derek

So how do drug molecules (and others) get into cells, anyway? There are two broad answers: they just sort of slide in through the membranes on their own (passive diffusion), or they're taken up by pores and proteins built for bringing things in (active transport). I've always been taught (and believed) that both processes can be operating in most situations. If the properties of your drug molecule stray too far out of the usual range, for example, your cell activity tends to drop, presumably because it's no longer diffusing past the cell membranes. There are other situations where you can prove that you're hitching a ride on active transport proteins, by administering a known inhibitor of one of these systems to cells and watching your compound suddenly become inactive, or by simply overloading and saturating the transporter.

There's another opinion, though, that's been advanced by Paul Dobson and Douglas Kell at Manchester, and co-workers. Their take is that carrier-mediated transport is the norm, and that passive diffusion is hardly important at all. This has been received with varying degrees of belief. Some people seem to find it a compelling idea, while others regard it as eccentric at best. The case was made a few years ago in Nature Reviews Drug Discovery, and again more recently in Drug Discovery Today:

All cells necessarily contain tens, if not hundreds, of carriers for nutrients and intermediary metabolites, and the human genome codes for more than 1000 carriers of various kinds. Here, we illustrate using a typical literature example the widespread but erroneous nature of the assumption that the ‘background’ or ‘passive’ permeability to drugs occurs in the absence of carriers. Comparison of the rate of drug transport in natural versus artificial membranes shows discrepancies in absolute magnitudes of 100-fold or more, with the carrier-containing cells showing the greater permeability. Expression profiling data show exactly which carriers are expressed in which tissues. The recognition that drugs necessarily require carriers for uptake into cells provides many opportunities for improving the effectiveness of the drug discovery process.

That's one of those death-or-glory statements: if it's right, a lot of us have been thinking about these things the wrong way, and missing out on some very important things about drug discovery as well. But is it? There's a rebuttal paper out in Drug Discovery Today that makes the case for the defense. It's by a long list of pharmacokinetics and pharmacology folks from industry and academia, and has the air of "Let's get this sorted out once and for all" about it:

Evidence supporting the action of passive diffusion and carrier-mediated (CM) transport in drug bioavailability and disposition is discussed to refute the recently proposed theory that drug transport is CM-only and that new transporters will be discovered that possess transport characteristics ascribed to passive diffusion. Misconceptions and faulty speculations are addressed to provide reliable guidance on choosing appropriate tools for drug design and optimization.

Fighting words! More of those occur in the body of the manuscript, phrases like "scientifically unsound", "potentially misleading", and "based on speculation rather than experimental evidence". Here's a rundown of the arguments, but if you don't read the paper, you'll miss the background noise of teeth being ground together.

Kell and Dobson et al. believe that cell membrane have more protein in them, and less lipid, than is commonly thought, which helps make their case for lots of protein transport/not a lot of lipid diffusion. But this paper says that their figures are incorrect and have been misinterpreted. Another K-D assertion is that artificial lipid membranes tend to have many transient aqueous pores in them, which make them look more permeable than they really are. This paper goes to some length to refute this, citing a good deal of prior art with examples of things which should have then crossed such membranes (but don't), and also find fault with the literature that K-D used to back up their own proposal.

This latest paper then goes on to show many examples of non-saturatable passive diffusion, as opposed to active transport, which can always be overloaded. Another big argument is over the agreement between different cell layer models of permeability. Two of the big ones are Caco-2 cells and MDCK cells, but (as all working medicinal chemists know) the permeability values between these two don't always agree, either with each other or with the situation in living systems. Kell and Dobson adduce this as showing the differences between the various transporters in these assays, but this rebuttal points out that there are a lot of experimental differences between literature Caco-2 and MDCK assays that can kick the numbers around. Their take is that the two assays actually agree pretty well, all things considered, and that if transporters were the end of the story that the numbers would be still farther apart.

The blood-brain barrier is a big point of contention between these two camps. This latest paper cites a large pile of literature showing that sheer physical properties (molecular weight, logP) account for most successful approaches to getting compounds into the brain, consistent with passive diffusion, while examples of using active transport are much more scarce. That leads into one of the biggest K-D points, which seems to be one of the ones that drives the existing pharmacokinetics community wildest: the assertion that thousands of transport proteins remain poorly characterized, and that these will come to be seen as the dominant players compared to passive mechanisms. The counterargument is that most of these, as far as we can tell to date, are selective for much smaller and more water-soluble substances than typical drug molecules (all the way from metal ions to things like glycerol and urea), and are unlikely to be important for most pharmaceuticals.

Relying on as-yet-uncharacterized transporters to save one's argument is a habit that really gets on the nerves of the Kell-Dobson critics as well - this paper calls it "pure speculation without scientific basis or evidence", which is about as nasty as we get in the technical literature. I invite interested readers to read both sides of the argument and make up their own minds. As for me, I fall about 80% toward the critics' side. I think that there are probably important transporters that are messing with our drug concentrations and that we haven't yet appreciated, but I just can't imagine that that's the whole story, nor that there's no such thing as passive diffusion. Thoughts?

Comments (37) + TrackBacks (0) | Category: Drug Assays | Pharma 101 | Pharmacokinetics

April 25, 2012

DHFR Inhibitors Revisited: A Word From the Authors (and Reviewers)

Email This Entry

Posted by Derek

The other day, I had some uncomplimentary things to say about a recent J. Med. Chem. paper on fragment-based dihydrofolate reductase inhibitors. Well, I know that I don't say these things into a vacuum, by any means, but in this case the lead author has written me about the work, and a reviewer of the paper has showed up in the comments. So perhaps this is a topic worth revisiting?

First, I'll give Prof. Joelle Pelletier of U. Montreal the floor to make the case for the defense. Links added are mine, for background; I take responsibility for those, and I hope they're helpful.

I was informed of your recent blog entitled ‘How do these things get published’. I am corresponding author of that paper. I would like to bring to your attention a crucial point that was incorrectly presented in your analysis: the target enzyme is not that which you think it is, i.e.: it is not a DHFR that is part of ‘a class of enzymes that's been worked on for decades’.

Indeed, it would make no sense to report weak and heavy inhibitors against ‘regular’ DHFRs (known as ‘type I DHFRs’), considering the number of efficient DHFR inhibitors we already know. But this target has no sequence or structural homology with type I DHFRs. It is a completely different protein that offers an alternate path to production of tetrahydrofolate (see top of second page of the article). It has apparently evolved recently, as a bacterial response to trimethoprim being introduced into the environment since the ‘60’s. Because that protein is evolutionarily unrelated to regular DHFRs, it doesn’t bind trimethoprim and is thus intrinsically trimethoprim resistant; it isn’t inhibited by other inhibitors of regular DHFRs either. There have been no efforts to date to inhibit this drug resistance enzyme, despite its increasing prevalence in clinical and veterinary settings, and in food and wastewater (see first page of article). As a result, we know nothing about how to prevent it from providing drug resistance. Our paper is thus the first foray into inhibiting this new target – one which presents both the beauty and the difficulty of complex symmetry.

Regular (type I) DHFRs are monomeric enzymes with an extended active-site cleft. They are chromosomally-encoded in all living cells where they are essential for cellular proliferation. Our target, type II R67 DHFR, is carried on a plasmid, allowing rapid dissemination between bacterial species. It is an unusual homotetrameric, doughnut-style enzyme with the particularity of having a single active site in the doughnut hole. That’s unusual because multimeric enzymes typically have the same number of active sites as they do monomers. The result is that the active site tunnel, shown in Figure 4 a, has 222 symmetry. Thus, the front and back entrances to the active site tunnel are identical. And that’s why designing long symmetrical molecules makes sense: they have the potential of threading through the tunnel, where the symmetry of the inhibitor would match the symmetry of the target. If they don’t string through but fold up into a ‘U”, it still makes sense: the top and bottom of the tunnel are also alike, again allowing a match-up of symmetry. Please note that this symmetry does create a bit of a crystallographer’s nightmare at the center of the tunnel where the axes of symmetry meet; again, it is an unusual system.

You have referred to our ‘small, poorly documented library of fragment compounds’. As for the poor documentation, the point is that we have very little prior information on the ligands of this new target, other than its substrates. We cast as wide a net as we could within a loosely defined chemical class, using the chemicals we have access to. Unfortunately, I don’t have access to a full fragment library, but am open to collaboration.

As a result of extending the fragments, the ligand efficiency does take a beating… so would it have been better not to mention it? No, that would have been dishonest. In addition, it is not a crucial point at this very early stage in discovery: this is a new target, and it IS important to obtain information on tighter binding, even if it comes at the cost of heavier molecules. In no way do we pretend that these molecules are ripe for application; we have presented the first set of crude inhibitors to ‘provide inspiration for the design of the next generation of inhibitors’ (last sentence of the paper).

Your blog is widely read and highly respected. In this case, it appears that your analysis was inaccurate due to a case of mistaken identity. I did appreciate your calm and rational tone, and hope that you will agree that there is redeeming value to the poor ligand efficiency, because of the inherent novelty of this discovery effort. I am appealing to you to reconsider the blog’s content in light of the above information, and respectfully request that you consider revising it.

Well, as for DHFRs, I'm guilty as charged. The bacterial ones really are way off the mammalian ones - it appears that dihydro/tetrahydrofolate metabolism is a problem that's been solved a number of different ways and (as is often the case) the bacteria show all kinds of diversity compared to the rest of the living world. And there really aren't any good D67 DHFR inhibitors out there, not selective ones, anyway, so a molecule of that type would definitely be a very worthwhile tool (as well as a potential antibiotic lead).

But that brings us to the fragments, the chemical matter in the paper. I'm going to stand my my characterization of the fragment library. 100 members is indeed small, and claiming lack of access to a "full fragment collection" doesn't quite cover it. Because of the amount of chemical space that can be covered at these molecular weights, a 200-member library can be significantly more useful than a 100-member one, and so on. (Almost anything is more useful than a 100-member library). There aren't more compounds of fragment size on the shelves at the University of Montreal?

More of a case could be made for libraries this small if they covered chemical space well. Unfortunately, looking over the list of compounds tested (which is indeed in the Supplementary Material), it's not, at first glance, a very good collection. Not at all. There are some serious problems, and in a collection this small, mistakes are magnified. I have to point out, to start with, that compounds #59 and #81 are duplicates, as are compounds #3 and #40, and compounds #7 and #14. (There may be others; I haven't made a complete check).

The collection is heavily biased towards carboxylic acids (which is a problem for several reasons, see below). Nearly half the compounds have a COOH group by my quick count, and it's not a good idea to have any binding motif so heavily represented. I realize that you intentionally biased your screening set, but then, an almost featureless hydrophobic compound like #46 has no business in there. Another problem is that some of the compounds are so small that they're unlikely to be tractable fragment hits - I note succinimide (#102) and propyleneurea (#28) as examples, but there are others. At the other end of the scale, compounds such as the Fmoc derivative #25 are too large (MW 373), and that's not the only offender in the group (nor the only Fmoc derivative). The body of the manuscript mentions the molecular weights of the collections as being from 150 to 250, but there are too many outliers. This isn't a large enough collection for this kind of noise to be in it.

There are a number of reactive compounds in the list, too, and while covalent inhibitors are a very interesting field, this was not mentioned as a focus of your efforts or as a component of the screening set. And even among these, compounds such as carbonyldiimidazole (#26), the isocyanate #82, and disuccinimidylcarbonate (#36) are really pushing it, as far as reactivity and hydrolytic stability. The imine #110 is also very small and likely to have hydrolytic stability problems. Finally, the fragment #101 is HEPES, which is rather odd, since HEPES is the buffer for the enzyme assays. Again, there isn't room for these kinds of mistakes. It's hard for me to imagine that anyone who's ever done fragment screening reviewed this manuscript.

The approach to following up these compounds also still appears inadequate to me. As Dan Erlanson pointed out in a comment to the Practical Fragments post, small carboxylic acids like the ones highlighted are not always legitimate hits. They can, as he says, form aggregates, depending on the assay conditions, and the most straightforward way of testing that is often the addition of a small amount of detergent, if the assay can stand it. The behavior of such compounds is also very pH-dependent, as I've had a chance to see myself on a fragment effort, so you need to make sure that you're as close to physiological conditions as you can get. I actually have seen some of your compounds show up as hits in fragment screening efforts, and they've been sometimes real, sometimes not.

But even if we stipulate that these compounds are actually hits, they need more work than they've been given. The best practice, in most cases when a fragment hit is discovered and confirmed, is to take as many closely related single-atom changes into the assay as possible. Scan a methyl group around the structure, scan a fluoro, make the N-for-C switches - at these molecular weights, these changes can make a big difference, and you may well find an even more ligand-efficient structure to work from.

Now, as for the SAR development that actually was done: I understand the point about the symmetry of the enzyme, and I can see why this led to the idea of making symmetrical dimer-type compounds. But, as you know, this isn't always a good idea. Doing so via flexible alkyl or alkyl ether chains is not a good idea, though, since such compounds will surely pay an entropic penalty in binding.

And here's one of the main things that struck both me and Teddy Z in his post: if the larger compounds were truly taking advantage of the symmetry, their ligand efficiency shouldn't go down. But in this case it does, and steeply. The size of the symmetical inhibitors (and their hydrophobic regions, such as the featureless linking chains, make it unsurprising that this effort found some micromolar activity. Lots of things will no doubt show micromolar activity in such chemical space. The paper notes that it's surprising that the fragment 4c showed no activity when its structural motif was used to build some of the more potent large compounds, but the most likely hypothesis is that this is because the binding modes have nothing to do with each other.

To be fair, compounds 8 and 9 are referred to as "poorly optimized", which is certainly true. But the paper goes on to say that they are starting points to develop potent and selective inhibitors, which they're not. The fragments are starting points, if they're really binding. The large compounds are dead ends. That's why Teddy Z and I have reacted as strongly as we have, because the path this paper takes is (to our eyes) an example of how not to do fragment-based drug discovery.

But still, I have to say that I'm very glad to hear a direct reply to my criticism of this paper. I hope that this exchange has been useful, and that it might be of use for others who read it.

Comments (24) + TrackBacks (0) | Category: Drug Assays | Infectious Diseases | The Scientific Literature

April 16, 2012

Phenotypic Screening's Comeback

Email This Entry

Posted by Derek

Here's an excellent overview of phenotypic screening at SciBx. For those outside the field, phenotypic screening is the way things used to be all the time in the drug discovery business, decades ago: (1) give compounds to a living system, and watch what happens. (2) Wait until you find a compound that does what you want, and develop that one if you can.

That's as opposed to target-based drug discovery, which began taking over in the 1970s or so, and has grown ever since as molecular biology advanced. That's where you figure out enough about a biochemical pathway to know what enzyme/receptor/etc. you should try to inhibit, and you screen against that one alone to find your leads. That has worked out very well in some cases, but not as often as people would have imagined back at the beginning.

In fact, I (and a number of other people) have been wondering if the whole molecular-biology target-based approach has been something of a dead end. A recent analysis suggested that phenotypic screens have been substantially more productive in generating first-in-class drugs, and an overemphasis on individual targets has been been suggested as a reason for the lack of productivity in drug discovery.

As that new article makes clear, though, in most cases of modern phenotypic screening, people are going back from their hit compounds and finding out how they work, when possible. That's actually an excellent platform for discoveries in biology, too, as well as for finding medicinally active compounds. I'm glad to see cell- and tissue-based assays making a comeback, and I hope that they can bail us all out a bit.

Comments (30) + TrackBacks (0) | Category: Drug Assays | Drug Industry History

April 6, 2012

Europe Wants Some of That Molecular Library Action

Email This Entry

Posted by Derek

We've talked about the NIH's Molecular Libraries Initiative here a few times, mostly in the context of whether it reached its goals, and what might happen now that it looks as if it might go away completely. Does make this item a little surprising?

Almost a decade ago, the US National Institutes of Health kicked off its Molecular Libraries Initiative to provide academic researchers with access to the high-throughput screening tools needed to identify new therapeutic compounds. Europe now seems keen on catching up.

Last month, the Innovative Medicines Initiative (IMI), a €2 billion ($2.6 billion) Brussels-based partnership between the European Commission and the European Federation of Pharmaceutical Industries and Associations (EFPIA), invited proposals to build a molecular screening facility for drug discovery in Europe that will combine the inquisitiveness of academic scientists with industry know-how. The IMI's call for tenders says the facility will counter “fragmentation” between these sectors.

I can definitely see the worth in that part of the initiative. Done properly, Screening Is Good. But they'll have to work carefully to make sure that their compound collection is worth screening, and to format the assays so that the results are worth looking at. Both those processes (library generation and high-throughput screening) are susceptible (are they ever) to "garbage in, garbage out" factors, and it's easy to kid yourself into thinking that you're doing something worthwhile just because you're staying so busy and you have so many compounds.

There's another part of this announcement that worries me a bit, though. Try this on for size:

Major pharmaceutical companies have more experience with high-throughput screening than do most academic institutes. Yet companies often limit tests of their closely held candidate chemicals to a fraction of potential disease targets. By pooling chemical libraries and screening against a more diverse set of targets—and identifying more molecular interactions—both academics and pharmaceutical companies stand to gain, says Hugh Laverty, an IMI project manager.

Well, sure, as I said above, Screening Is Good, when it's done right, and we do indeed stand to learn things we didn't know before. But is it really true that we in the industry only look at a "fraction of potential disease targets"? This sounds like someone who's keen to go after a lot of the tough ones; the protein-protein interactions, protein-nucleic acid interactions, and even further afield. Actually, I'd encourage these people to go for it - but with eyes open and brain engaged. The reason that we don't screen against such things as often is that hit rates tend to be very, very low, and even those are full of false positives and noise. In fact, for many of these things, "very, very low" is not distinguishable from "zero". Of course, in theory you just need one good hit, which is why I'm still encouraging people to take a crack. But you should do so knowing the odds, and be ready to give your results some serious scrutiny. If you think that there must be thousands of great things out there that the drug companies are just too lazy (or blinded by the thought of quick profits elsewhere) to pursue, you're not thinking this through well enough.

You might say that what these efforts are looking for are tool compounds, not drug candidates. And I think that's fine; tool compounds are valuable. But if you read that news link in the first paragraph, you'll see that they're already talking about how to manage milestone payments and the like. That makes me think that someone, at any rate, is imagining finding valuable drug candidates from this effort. The problem with that is that if you're screening all the thousands of drug targets that the companies are ignoring, you're by definition working with targets that aren't very validated. So any hits that you do find (and there may not be many, as said above) will still be against something that has a lot of work yet to be done on it. It's a bit early to be wondering how to distribute the cash rewards.

And if you're screening against validated targets, the set of those that don't have any good chemical matter against them already is smaller (and it's smaller for a reason). It's not that there aren't any, though: I'd nominate PTP1B as a well-defined enzymatic target that's just waiting for a good inhibitor to come along to see if it performs as well in humans as it does in, say, knockout mice. (It's both a metabolic target and a potential cancer target as well). Various compounds have been advanced over the years, but it's safe to say that they've been (for the most part) quite ugly and not as selective as they could have been. People are still whacking away at the target.

So any insight into decent-looking selective phosphatase inhibitors would be most welcome. And most unlikely, damn it all, but all great drug ideas are most unlikely. The people putting this initiative together will have a lot to balance.

Comments (20) + TrackBacks (0) | Category: Academia (vs. Industry) | Biological News | Drug Assays

April 4, 2012

The Artificial Intelligence Economy?

Email This Entry

Posted by Derek

Now here's something that might be about to remake the economy, or (on the other robotic hand) it might not be ready to just yet. And it might be able to help us out in drug R&D, or it might turn out to be mostly beside the point. What the heck am I talking about, you ask? The so-called "Artificial Intelligence Economy". As Adam Ozimek says, things are looking a little more futuristic lately.

He's talking about things like driverless cars and quadrotors, and Tyler Cowen adds the examples of things like Apple's Siri and IBM's Watson, as part of a wider point about American exports:

First, artificial intelligence and computing power are the future, or even the present, for much of manufacturing. It’s not just the robots; look at the hundreds of computers and software-driven devices embedded in a new car. Factory floors these days are nearly empty of people because software-driven machines are doing most of the work. The factory has been reinvented as a quiet place. There is now a joke that “a modern textile mill employs only a man and a dog—the man to feed the dog, and the dog to keep the man away from the machines.”

The next steps in the artificial intelligence revolution, as manifested most publicly through systems like Deep Blue, Watson and Siri, will revolutionize production in one sector after another. Computing power solves more problems each year, including manufacturing problems.

Two MIT professors have written a book called Race Against the Machine about all this, and it appears to be sort of a response to Cowen's earlier book The Great Stagnation. (Here's an article of theirs in The Atlantic making their case).

One of the export-economy factors that it (and Cowen) bring up is that automation makes a country's wages (and labor costs in general) less of a factor in exports, once you get past the capital expenditure. And as the size of that expenditure comes down, it becomes easier to make that leap. One thing that means, of course, is that less-skilled workers find it harder to fit in. Here's another Atlantic article, from the print magazine, which looked at an auto-parts manufacturer with a factory in South Carolina (the whole thing is well worth reading):

Before the rise of computer-run machines, factories needed people at every step of production, from the most routine to the most complex. The Gildemeister (machine), for example, automatically performs a series of operations that previously would have required several machines—each with its own operator. It’s relatively easy to train a newcomer to run a simple, single-step machine. Newcomers with no training could start out working the simplest and then gradually learn others. Eventually, with that on-the-job training, some workers could become higher-paid supervisors, overseeing the entire operation. This kind of knowledge could be acquired only on the job; few people went to school to learn how to work in a factory.
Today, the Gildemeisters and their ilk eliminate the need for many of those machines and, therefore, the workers who ran them. Skilled workers now are required only to do what computers can’t do (at least not yet): use their human judgment.

But as that article shows, more than half the workers in that particular factory are, in fact, rather unskilled, and they make a lot more than their Chinese counterparts do. What keeps them employed? That calculation on what it would take to replace them with a machine. The article focuses on one of those workers in particular, named Maddie:

It feels cruel to point out all the Level-2 concepts Maddie doesn’t know, although Maddie is quite open about these shortcomings. She doesn’t know the computer-programming language that runs the machines she operates; in fact, she was surprised to learn they are run by a specialized computer language. She doesn’t know trigonometry or calculus, and she’s never studied the properties of cutting tools or metals. She doesn’t know how to maintain a tolerance of 0.25 microns, or what tolerance means in this context, or what a micron is.

Tony explains that Maddie has a job for two reasons. First, when it comes to making fuel injectors, the company saves money and minimizes product damage by having both the precision and non-precision work done in the same place. Even if Mexican or Chinese workers could do Maddie’s job more cheaply, shipping fragile, half-finished parts to another country for processing would make no sense. Second, Maddie is cheaper than a machine. It would be easy to buy a robotic arm that could take injector bodies and caps from a tray and place them precisely in a laser welder. Yet Standard would have to invest about $100,000 on the arm and a conveyance machine to bring parts to the welder and send them on to the next station. As is common in factories, Standard invests only in machinery that will earn back its cost within two years. For Tony, it’s simple: Maddie makes less in two years than the machine would cost, so her job is safe—for now. If the robotic machines become a little cheaper, or if demand for fuel injectors goes up and Standard starts running three shifts, then investing in those robots might make sense.

At this point, some similarities to the drug discovery business will be occurring to readers of this blog, along with some differences. The automation angle isn't as important, or not yet. While pharma most definitely has a manufacturing component (and how), the research end of the business doesn't resemble it very much, despite numerous attempts by earnest consultants and managers to make it so. From an auto-parts standpoint, there's little or no standardization at all in drug R&D. Every new drug is like a completely new part that no one's ever built before; we're not turning out fuel injectors or alternators. Everyone knows how a car works. Making a fundamental change in that plan is a monumental challenge, so the auto-parts business is mostly about making small variations on known components to the standards of a given customer. But in pharma - discovery pharma, not the generic companies - we're wrenching new stuff right out of thin air, or trying to.

So you'd think that we wouldn't be feeling the low-wage competitive pressure so much, but as the last ten years have shown, we certainly are. Outsourcing has come up many a time around here, and the very fact that it exists shows that not all of drug research is quite as bespoke as we might think. (Remember, the first wave of outsourcing, which is still very much a part of the business, was the move to send the routine methyl-ethyl-butyl-futile analoging out somewhere cheaper). And this takes us, eventually, to the Pfizer-style split between drug designers (high-wage folks over here) and the drug synthesizers (low-wage folks over there). Unfortunately, I think that you have to go the full reducio ad absurdum route to get that far, but Pfizer's going to find out for us if that's an accurate reading.

What these economists are also talking about is, I'd say, the next step beyond Moore's Law: once we have all this processing power, how do we use it? The first wave of computation-driven change happened because of the easy answers to that question: we had a lot of number-crunching that was being done by hand, or very slowly by some route, and we now had machines that could do what we wanted to do more quickly. This newer wave, if wave it is, will be driven more by software taking advantage of the hardware power that we've been able to produce.

The first wave didn't revolutionize drug discovery in the way that some people were hoping for. Sheer brute force computational ability is of limited use in drug discovery, unfortunately, but that's not always going to be the case, especially as we slowly learn how to apply it. If we really are starting to get better at computational pattern recognition and decision-making algorithms, where could that have an impact?

It's important to avoid what I've termed the "Andy Grove fallacy" in thinking about all this. I think that it is a result of applying first-computational-wave thinking too indiscriminately to drug discovery, which means treating it too much like a well-worked-out human-designed engineering process. Which it certainly isn't. But this second-wave stuff might be more useful.

I can think of a few areas: in early drug discovery, we could use help teasing patterns out of large piles of structure-activity relationship data. I know that there are (and have been) several attempts at doing this, but it's going to be interesting to see if we can do it better. I would love to be able to dump a big pile of structures and assay data points into a program and have it say the equivalent of "Hey, it looks like an electron-withdrawing group in the piperidine series might be really good, because of its conformational similarity to the initial lead series, but no one's ever gotten back around to making one of those because everyone got side-tracked by the potency of the chiral amides".

Software that chews through stacks of PK and metabolic stability data would be worth having, too, because there sure is a lot of it. There are correlations in there that we really need to know about, that could have direct relevance to clinical trials, but I worry that we're still missing some of them. And clinical trial data itself is the most obvious place for software that can dig through huge piles of numbers, because those are the biggest we've got. From my perspective, though, it's almost too late for insights at that point; you've already been spending the big money just to get the numbers themselves. But insights into human toxicology from all that clinical data, that stuff could be gold. I worry that it's been like the concentration of gold in seawater, though: really there, but not practical to extract. Could we change that?

All this makes me actually a bit hopeful about experiments like this one that I described here recently. Our ignorance about medicine and human biochemistry is truly spectacular, and we need all the help we can get in understanding it. There have to be a lot of important things out there that we just don't understand, or haven't even realized the existence of. That lack of knowledge is what gives me hope, actually. If we'd already learned what there is to know about discovering drugs, and were already doing the best job that could be done, well, we'd be in a hell of a fix, wouldn't we? But we don't know much, we're not doing it as well as we could, and that provides us with a possible way out of the fix we're in.

So I want to see as much progress as possible in the current pattern-recognition and data-correlation driven artificial intelligence field. We discovery scientists are not going to automate ourselves out of business so quickly as factory workers, because our work is still so hypothesis-driven and hard to define. (For a dissenting view, with relevance to this whole discussion, see here). It's the expense of applying the scientific method to human health that's squeezing us all, instead, and if there's some help available in that department, then let's have it as soon as possible.

Comments (32) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History | In Silico | Pharmacokinetics | Toxicology

March 29, 2012

Sloppy Science

Email This Entry

Posted by Derek

Nature has a comment on the quality of recent publications in clinical oncology. And it's not a kind one:

Glenn Begley and Lee Ellis analyse the low number of cancer-research studies that have been converted into clinical success, and conclude that a major factor is the overall poor quality of published preclinical data. A warning sign, they say, should be the “shocking” number of research papers in the field for which the main findings could not be reproduced. To be clear, this is not fraud — and there can be legitimate technical reasons why basic research findings do not stand up in clinical work. But the overall impression the article leaves is of insufficient thoroughness in the way that too many researchers present their data.

The finding resonates with a growing sense of unease among specialist editors on this journal, and not just in the field of oncology. Across the life sciences, handling corrections that have arisen from avoidable errors in manuscripts has become an uncomfortable part of the publishing process.

I think that this problem has been with us for quite a while, and that there are a few factors making it more noticeable: more journals to publish in, for one thing, and increased publication pressure, for another. And the online availability of papers makes it easier to compare publications and to call them up quickly; things don't sit on the shelf in quite the way that they used to. But there's no doubt that a lot of putatively interesting results in the literature are not real. To go along with that link, the Nature article itself referred to in that commentary has some more data:

Over the past decade, before pursuing a particular line of research, scientists. . .in the haematology and oncology department at the biotechnology firm Amgen in Thousand Oaks, California, tried to confirm published findings related to that work. Fifty-three papers were deemed 'landmark' studies. . . It was acknowledged from the outset that some of the data might not hold up, because papers were deliberately selected that described something completely new, such as fresh approaches to targeting cancers or alternative clinical uses for existing therapeutics. Nevertheless, scientific findings were confirmed in only 6 (11%) cases. Even knowing the limitations of preclinical research, this was a shocking result.

Of course, the validation attempts may have failed because of technical differences or difficulties, despite efforts to ensure that this was not the case. Additional models were also used in the validation, because to drive a drug-development programme it is essential that findings are sufficiently robust and applicable beyond the one narrow experimental model that may have been enough for publication. To address these concerns, when findings could not be reproduced, an attempt was made to contact the original authors, discuss the discrepant findings, exchange reagents and repeat experiments under the authors' direction, occasionally even in the laboratory of the original investigator. These investigators were all competent, well-meaning scientists who truly wanted to make advances in cancer research.

So what leads to these things not working out? Often, it's trying to run with a hypothesis, and taking things faster than they can be taken:

In studies for which findings could be reproduced, authors had paid close attention to controls, reagents, investigator bias and describing the complete data set. For results that could not be reproduced, however, data were not routinely analysed by investigators blinded to the experimental versus control groups. Investigators frequently presented the results of one experiment, such as a single Western-blot analysis. They sometimes said they presented specific experiments that supported their underlying hypothesis, but that were not reflective of the entire data set. . .

This can rise, on occasion, to the level of fraud, but it's not fraud if you're fooling yourself, too. Science is done by humans, and it's always going to have a fair amount of slop in it. The same issue of Nature, as fate would have it has a good example of irreproducibility this week. Sanofi's PARP inhibitor iniparib already wiped out in Phase III clinical trials not long ago, after having looked good in Phase II. It now looks as if the compound was (earlier reports notwithstanding) never much of a PARP1 inhibitor at all. (Since one of these papers is from Abbott, you can see that doubts had already arisen elsewhere in the industry).

That's not the whole story with PARP - AstraZeneca had a real inhibitor, olaparib, fail on them recently, so there may well be a problem with the whole idea. But iniparib's mechanism-of-action problems certainly didn't help to clear anything up.

Begley and Ellis call for tightening up preclinical oncology research. There are plenty of cell experiments that will not support the claims made for them, for one thing, and we should stop pretending that they do. They also would like to see blinded protocols followed, even preclinically, to try to eliminate wishful thinking. That's a tall order, but it doesn't mean that we shouldn't try.

Update: here's more on the story. Try this quote:

Part way through his project to reproduce promising studies, Begley met for breakfast at a cancer conference with the lead scientist of one of the problematic studies.

"We went through the paper line by line, figure by figure," said Begley. "I explained that we re-did their experiment 50 times and never got their result. He said they'd done it six times and got this result once, but put it in the paper because it made the best story. It's very disillusioning."

Comments (38) + TrackBacks (0) | Category: Cancer | Drug Assays | The Scientific Literature

March 19, 2012

Dealing with the Data

Email This Entry

Posted by Derek

So how do we deal with the piles of data? A reader sent along this question, and it's worth thinking about. Drug research - even the preclinical kind - generates an awful lot of information. The other day, it was pointed out that one of our projects, if you expanded everything out, would be displayed on a spreadsheet with compounds running down the left, and over two hundred columns stretching across the page. Not all of those are populated for every compound, by any means, especially the newer ones. But compounds that stay in the screening collection tend to accumulate a lot of data with time, and there are hundreds of thousands (or millions) of compounds in a good-sized screening collection. How do we keep track of it all?

Most larger companies have some sort of proprietary software for the job (or jobs). The idea is that you can enter a structure (or substructure) of a compound and find out the project it was made for, every assay that's been run on it, all its spectral data and physical properties (experimental and calculated), every batch that's been made or bought (and from whom and from where, with notebook and catalog references), and the bar code of every vial or bottle of it that's running around the labs. You obviously don't want all of those every time, so you need to be able to define your queries over a wide range, setting a few common ones as defaults and customizing them for individual projects while they're running.

Displaying all this data isn't trivial, either. The good old fashioned spreadsheet is perfectly useful, but you're going to need the ability to plot and chart in all sorts of ways to actually see what's going on in a big project. How does human microsomal stability relate to the logP of the right-hand side chain in the pyrimidinyl-series compounds with molecular weight under 425? And how do those numbers compare to the dog microsomes? And how do either of those compare to the blood levels in the whole animal, keeping in mind that you've been using two different dosing vehicles along the way? To visualize these kinds of questions - perfectly reasonable ones, let me tell you - you'll need all the help you can get.

You run into the problem of any large, multifunctional program, though: if it can do everything, it may not do any one thing very well. Or there may be a way to do whatever you want, if only you can memorize the magic spell that will make it happen. If it's one of those programs that you have to use constantly or run the risk of totally forgetting how it goes, there will be trouble.

So what's been the experience out there? In-house home-built software? Adaptations of commercial packages? How does a smaller company afford to do what it needs to do? Comments welcome. . .

Comments (66) + TrackBacks (0) | Category: Drug Assays | Drug Development | Life in the Drug Labs

March 14, 2012

The Blackian Demon of Drug Discovery

Email This Entry

Posted by Derek

There's an on-line appendix to that Nature Reviews Drug Discovery article that I've been writing about, and I don't think that many people have read it yet. Jack Scannell, one of the authors, sent along a note about it, and he's interested to see what the readership here makes of it.

It gets to the point that came up in the comments to this post, about the order that you do your screening assays in (see #55 and #56). Do you run everything through a binding assay first, or do you run things through a phenotypic assay first and then try to figure out how they bind? More generally, with either sort of assay, is it better to do a large random screen first off, or is it better to do iterative rounds of SAR from a smaller data set? (I'm distinguishing those two because phenotypic assays provide very different sorts of data density than do focused binding assays).

Statistically, there's actually a pretty big difference there. I'll quote from the appendix:

Imagine that you know all of the 600,000 or so words in the English language and that you are asked to guess an English word written in a sealed envelope. You are offered two search strategies. The first is the familiar ‘20 questions’ game. You can ask a series of questions. You are provided with a "yes" or "no" answer to each, and you win if you guess the word in the envelope having asked 20 questions or fewer. The second strategy is a brute force method. You get 20,000 guesses, but you only get a "yes" or "no" once you have made all 20,000 guesses. So which is more likely to succeed, 20 questions or 20,000 guesses?

A skilled player should usually succeed with 20 questions (since 600,000 is less than than 2^20) but would fail nearly 97% of the time with "only" 20,000 guesses.

Our view is that the old iterative method of drug discovery was more like 20 questions, while HTS of a static compound library is more like 20,000 guesses. With the iterative approach, the characteristics of each molecule could be measured on several dimensions (for example, potency, toxicity, ADME). This led to multidimensional structure–activity relationships, which in turn meant that each new generation of candidates tended to be better than the previous generation. In conventional HTS, on the other hand, search is focused on a small and pre-defined part of chemical space, with potency alone as the dominant factor for molecular selection.

Aha, you say, but the game of twenty questions is equivalent to running perfect experiments each time: "Is the word a noun? Does it have more than five letters?" and so on. Each question carves up the 600,000 word set flawlessly and iteratively, and you never have to backtrack. Good experimental design aspires to that, but it's a hard standard to reach. Too often, we get answers that would correspond to "Well, it can be used like a noun on Tuesdays, but if it's more than five letters, then that switches to Wednesday, unless it starts with a vowel".

The authors try to address this multi-dimensionality with a thought experiment. Imagine chemical SAR space - huge number of points, large number of parameters needed to describe each point.

Imagine we have two search strategies to find the single best molecule in this space. One is a brute force search, which assays a molecule and then simply steps to the next molecule, and so exhaustively searches the entire space. We call this "super-HTS". The other, which we call the “Blackian demon” (in reference to the “Darwinian demon”, which is used sometimes to reflect ideal performance in evolutionary thought experiments, and in tribute to James Black, often acknowledged as one of the most successful drug discoverers), is equivalent to an omniscient drug designer who can assay a molecule, and then make a single chemical modification to step it one position through chemical space, and who can then assay the new molecule, modify it again, and so on. The Blackian demon can make only one step at a time, to a nearest neighbour molecule, but it always steps in the right direction; towards the best molecule in the space. . .

The number of steps for the Blackian demon follows from simple geometry. If you have a d dimensional space with n nodes in the space, and – for simplicity – these are arranged in a neat line, square, cube, or hypercube, you can traverse the entire space, from corner to corner with d x (n^(1/d)-1) steps. This is because each vertex is n nodes in length, and there are d vertices. . .When the search space is high dimensional (as is chemical space) and there is a very large number of nodes (as is the case for drug-like molecules), the Blackian demon is many orders of magnitude more efficient than super-HTS. For example, in a 10 dimensional space with 10^40 molecules, the Blackian demon can search the entire space in 10^5 steps (or less), while the brute force method requires 10^40 steps.

These are idealized cases, needless to say. One problem is that none of us are exactly Blackian demons - what if you don't always make the right step to the next molecule? What if your iteration only gives one out of ten molecules that get better, or one out of a hundred? I'd be interested to see how that affects the mathematical argument.

And there's another conceptual problem: for many points in chemical space, the numbers are even much more sparse. One assumption with this thought experiment (correct me if I'm wrong) is that there actually is a better node to move to each time. But for any drug target, there are huge regions of flat, dead, inactive, un-assayable chemical space. If you started off in one of those, you could iterate until your hair fell out and never get out of the hole. And that leads to another objection to the ground rules of this exercise: no one tries to optimize by random HTS. It's only used to get starting points for medicinal chemists to work on, to make sure that they're not starting in one of those "dead zones". Thoughts?

Comments (45) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

March 6, 2012

Drug Discovery for Physicists

Email This Entry

Posted by Derek

There's a good post over at the Curious Wavefunction on the differences between drug discovery and the more rigorous sciences. I particularly liked this line:

The goal of many physicists was, and still is, to find three laws that account for at least 99% of the universe. But the situation in drug discovery is more akin to the situation in finance described by the physicist-turned-financial modeler Emanuel Derman; we drug hunters would consider ourselves lucky to find 99 laws that describe 3% of the drug discovery universe.

That's one of the things that you get used to in this field, but when you step back, it's remarkable: so much of what we do remains relentlessly empirical. I don't just mean finding a hit in a screening assay. It goes all the way through the process, and the further you go, the more empirical it gets. Cell assays surprise you compared to enzyme preps, and animals are a totally different thing than cells. Human clinical trials are the ultimate in empirical data-gathering: there's no other way to see if a drug is truly safe (or effective) in humans other than giving it to a whole big group of humans. We do all sorts of assays to avoid getting to that stage, or to feel more confident when we're about to make it there, but there's no substituted for actually doing it.

There's a large point about reductionism to be made, too:

Part of the reason drug discovery can be challenging to physicists is because they are steeped in a culture of reductionism. Reductionism is the great legacy of twentieth-century physics, but while it worked spectacularly well for particle physics it doesn't quite work for drug design. A physicist may see the human body or even a protein-drug system as a complex machine whose understandings we can completely understand once we break it down into its constituent parts. But the chemical and biological systems that drug discoverers deal with are classic examples of emergent phenomena. A network of proteins displays properties that are not obvious from the behavior of the individual proteins. . .Reductionism certainly doesn't work in drug discovery in practice since the systems are so horrendously complicated, but it may not even work in principle.

And there we have one of the big underlying issues that needs to be faced by the hardware engineers, software programmers, and others who come in asking why we can't be as productive as they are. There's not a lot of algorithmic compressibility in this business. Whether they know it or not, many other scientists and engineers are living in worlds where they're used to it being there when they need it. But you won't find much here.

Comments (22) + TrackBacks (0) | Category: Drug Assays | Drug Development

February 28, 2012

More on the NIH's Molecular Libraries Program

Email This Entry

Posted by Derek

I last wrote about the Molecular Libraries program here, as it was threatened with funding cuts. Now there's a good roundup of opinion on it here, at the SLAS. The author has looked over the thoughts of the readership here, and also heard from several other relevant figures. Chris Lipinski echoes what several commenters here had to say:

Lipinski notes that when the screening library collection began the NIH had little medicinal chemistry experience. "I was a member of an early teleconference to discuss what types of compounds should be acquired by the NIH for high-throughput screening (HTS) to discover chemical biology tools and probes. Our teleconference group was about evenly split between industry people and academics. The academics talked about innovation, thinking out of the box, maximum chemical diversity and not being limited by preconceived rules and filters. The industry people talked about pragmatism, the lessons learned and about worthless compounds that could appear active in HTS screens. The NIH was faced with two irreconcilable viewpoints. They had to pick one and they chose the academic viewpoint."

He says that they later moved away from this, with more success, but implies that quite a bit of time was lost before this happened. Now, we waste plenty of time and money in the drug industry, so I have no standing to get upset with the NIH about blind alleys, in principle. But having them waste time and money specifically on something that the drug industry could have warned them off of is another thing.

In the end, opinions divide (pretty much as you'd guess) on the worth of the whole initiative. As that link shows, its director believes it to have been a great success, while others give it more mixed reviews. Its worth has surely grown with time, though, as some earlier mistakes were corrected, and that's what seems to be worrying people: that the plug is getting pulled just when things were becoming more useful. It seems certain that several of the screening centers will not survive in the current funding environment. And what happens to their compounds then?

Comments (11) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays

February 10, 2012

The Infinitely Active Impurity

Email This Entry

Posted by Derek

Everyone who's done drug discovery has encountered this situation: you get what looks like a hit in a screening assay, but when you re-check it with fresh material, it turns out to be inactive. So you go back to the original batch, but it's still active. There are several possibilities: if that original batch was a DMSO solution, perhaps the compound has done something funny on standing, and you don't have what you thought you had. Maybe the DMSO stock was made from the wrong compound, or was mislabeled somehow - in which case, good luck figuring out what's really in there. If the original batch was a solid, the first thing to do is a head-to-head analysis (NMR, LC-mass spec) between the two. (That sort of purity check is actually the first thing you should do with interesting screening hits in general, as experienced chemists will have had several chances to learn).

But if those assay numbers repeat for both batches, you're in the realm of the Infinitely Active Impurity. The thinking is, and it's hard to find fault with it, that there must be something in Batch One that's causing the assay to light up, something that's not present in Batch Two. I found myself in this situation one time where the problem turned out to be that Batch One had the right structure, except it was a zinc complex, a fact the original submitters apparently hadn't appreciated. (We had to send out for metals analysis to confirm that one). In that case, the assay could be made to show a hit by adding zinc to most any compound you wanted, which wasn't too useful.

Most of the time, chasing after these things proves futile, which is frustrating for everyone involved. But not always. There's a recent example of a successful impurity hunt in ACS Medicinal Chemistry Letters, from a group at Pfizer searching for inhibitors of kynurenine aminotransferase II.
One of the hits was that compound 6 shown in the figure, but a second batch of it showed no activity at all. They dug into the original sample, and found that there was a touch of the N-hydroxy compound in it, and that was the reason for all the activity. Interestingly, it turns out that the amino group was involved in a covalent interaction with the enzyme's cofactor, pyridoxal-5′-phosphate (PLP). That's one of the things you probably want to suspect when you find such tiny amounts of a compound having such a large effect.

It's not a deal-breaker, but it's something to keep in mind. The whole topic of irreversible inhibitors has come up around here before, but it's worth another post soon, in light of the recent acquisition of Avila Pharmaceuticals, who specialized in this field. In this case, the compound isn't covalently attached to the protein, but rather to its bound cofactor, which would make people breath a bit easier. (And the group responsible for the covalency, an amine, isn't something to worry about, either).

Still, it's interesting to see this part of the paper:

"Although irreversible inhibition was not one of our lead criteria at the outset of the program, maintaining this attribute of 7 was a high priority through our optimization efforts. The potential advantages of irreversible inhibitors include low dose requirements and reduced off-target toxicity."

I say that because increased off-target toxicity has always been the worry with covalent drugs. But there's been a real revival of interest in the last few years - more on this next week.

Comments (16) + TrackBacks (0) | Category: Drug Assays | The Central Nervous System

February 2, 2012

Fluorine NMR: Why Not?

Email This Entry

Posted by Derek

Fluorine NMR is underused in chemistry. Well, then again, maybe it's not, but it's one of those thing that just seems like it should have more uses than it does. (Here's a recent bookon the subject). Fluorine is a great NMR nucleus - all the F in the world is the same isotope, unless you're right next to a PET scanning facility - and the different compound show up over a very wide range of chemical shifts. You've got that going for you, coupling information, NOE, basically all your friends from proton NMR.

There's a pretty recent paper showing a good use of all these qualities (blogged about here at Practical Fragments as well). A group at Amgen reports on their work using fluorine NMR as a fragment-screening tool. They can take mixtures of 10 or 12 compounds at a time (because of all those different chemical shifts) and run the spectra with and without a target protein in the vial. If a fragment binds, its F peak broadens out (you can even get binding constants if you run at a few different concentrations). A simple overlay of the two spectra tells you immediately if you have hits. You don't need to have any special form of the protein, and you don't even need to run in deuterated solvents, since you're just ignoring protons altogether.

Interestingly, when they go on to try other assay techniques as follow-up, they find that the fluorines themselves aren't always a key part of the binding. Sometimes switching to the non-fluorinated version of the fragment gives you a better compound; sometimes it doesn't. The binding constants you get from the NMR, though, do compare very well to the ones from other assays.

The part I found most interesting was the intra-ligand NOE example. (That's also something that's been done in proton NMR, although it's not easy). They show a case where 19F ligands do get close enough to show the effect, and that a linked version of the two fragments does, as you'd hope, make a much more potent compound. That's the sort of thing that fragment people are always wanting to know - what fits next door to my hit? Can they be linked together? Fragment linking has its ups and downs, going back to the Abbott SAR-by-NMR days. That was a technique that never really panned out, as far as can be seen, but this is at least an experimentally easy way to give it a shot. (Of course, the chances of the fluorines on your ligands actually being pointed at each other is probably small, so that does cancel things out a bit).

Overall, it's a fun paper to read - well, allowing for my geeky interests, it is - and perhaps it'll lead a few more people to think of things that could be done with fluorine NMR in general. It's just sitting there, waiting to be used. . .

Comments (9) + TrackBacks (0) | Category: Analytical Chemistry | Drug Assays

January 6, 2012

Do We Believe These Things, Or Not?

Email This Entry

Posted by Derek

Some of the discussions that come up here around clinical attrition rates and compound properties prompts me to see how much we can agree on. So, are these propositions controversial, or not?

1. Too many drugs fail in clinical trials. We are having a great deal of trouble going on with these failure rates, given the expense involved.

2. A significant number of these failures are due to lack of efficacy - either none at all, or not enough.

2a. Fixing efficacy failures is hard, since it seems to require deeper knowledge, case-by-case, of disease mechanisms. As it stands, we get a significant amount of this knowledge from our drug failures themselves.

2b. Better target selection without such detailed knowledge is hard to come by. Good phenotypic assays are perhaps the only shortcut, but a good phenotypic assays are not easy to develop and validate.

3. Outside of efficacy, a significant number of clinical failures are also due to side effects/toxicity. These two factors (efficacy and tox) account for the great majority of compounds that drop out of the clinic.

3a. Fixing tox/side effect failures through detailed knowledge is perhaps hardest of all, since there are a huge number of possible mechanisms. There are far more ways for things to go wrong than there are for them to work correctly.

3b. But there are broad correlations between molecular structures and properties and the likelihood of toxicity. While not infallible, these correlations are strong enough to be useful, and we should be grateful for anything we can get that might diminish the possibility of later failure.

Example of such structural features are redox-active groups like nitros and quinones, which really are associated with trouble - not invariably, but enough to make you very cautious. More broadly, high logP values are also associated with trouble in development - not as strongly, but strong enough to be worth considering.

So, is everyone pretty much in agreement with these things? What I'm saying is that if you take a hundred aryl nitro compounds into development, versus a hundred that don't have such a group, the latter cohort of compounds will surely have a higher success rate. And if you take a hundred compounds with logP values of 1 to 3 into development, these will have a higher success rate than a hundred compounds, against the same targets, with logP of 4 to 6. Do we believe this, or not?

Comments (34) + TrackBacks (0) | Category: Drug Assays | Drug Development | Toxicology

January 5, 2012

Lead-Oriented Synthesis - What Might That Be?

Email This Entry

Posted by Derek

A new paper in Angewandte Chemie tries to open another front in relations between academic and drug industry chemists. It's from several authors at GSK-Stevenage, and it proposes something they're calling "Lead-Oriented Synthesis". So what's that?

Well, the paper itself starts out as a quick tutorial on the state and practice of medicinal chemistry. That's a good plan, since Angewandte Chemie is not primarily a med-chem journal (he said with a straight face). Actually, it has the opposite reputation, a forum where high-end academic chemistry gets showplaced. So the authors start off by reminded the readership what drug discovery entails. And although we've had plenty of discussions around here about these topics, I think that most people can agree on the main points laid out:

1. Physical properties influence a drug's behavior.
2. Among those properties, logP may well be the most important single descriptor,
3. Most successful drugs have logP values between 1 and perhaps 4 or 5. Pushing the lipophilicity end of things is, generally speaking, asking for trouble.
4. Since optimization of lead compounds almost always adds molecular weight, and very frequently adds lipophilicity, lead compounds are better found in (and past) the low ends of these property ranges, to reduce the risk of making an unwieldy final compound.

As the authors take pains to say, though, there are many successful drugs that fall outside these ranges. But many of those turn out to have some special features - antibacterial compounds (for example) tend to be more polar outliers, for reasons that are still being debated. There is, though, no similar class of successful less polar than usual drugs, to my knowledge. If you're starting a program against a target that you have no reason to think is an outlier, and assuming you want an oral drug for it, then your chances for success do seem to be higher within the known property ranges.

So, overall, the GSK folks maintain that lead compounds for drug discovery are most desirable with logP values between -1 and 3, molecular weights from around 200 to 350, and no problematic functional groups (redox-active and so on). And I have to agree; given the choice, that's where I'd like to start, too. So why are they telling all this to the readers of Angewandte Chemie? Because these aren't the sorts of compounds that academic chemists are interested in making.

For example, a survey of the 2009 issues of the Journal of Organic Chemistry found about 32,700 compounds indexed with the word "preparation" in Chemical Abstracts, after organometallics, isotopically labeled compounds, and commercially available ones were stripped out. 60% of those are outside the molecular weight criteria for lead-like compounds. Over half the remainder fail cLogP, and most of the remaining ones fail the internal GSK structural filters for problematic functional groups. Overall, only about 2% of the JOC compounds from that year would be called "lead-like". A similar analysis across seven other synthetic organic journals led to almost the same results.

Looking at array/library synthesis, as reported in the Journal of Combinatorial Chemistry and from inside GSK's own labs, the authors quantify something else that most chemists suspected: the more polar structures tend to drop out as the work goes on. This "cLogP drift" seems to be due to incompatible chemistries or difficulties in isolation and purification, and this could also illustrate why many new synthetic methods aren't applied in lead-like chemical space: they don't work as well there.

So that's what underlies the call for "lead-oriented synthesis". This paper is asking for the development of robust reactions which will work across a variety of structural types, will be tolerant of polar functionalities, and will generate compounds without such potentially problematic groups as Michael acceptors, nitros, and the like. That's not so easy, when you actually try to do it, and the hope is that it's enough of a challenge to attract people who are trying to develop new chemistry.

Just getting a high-profile paper of this sort out into the literature could help, because it's something to reference in (say) grant applications, to show that the proposed research is really filling a need. Academic chemists tend, broadly, to work on what will advance or maintain their positions and careers, and if coming up with new reactions of this kind can be seen as doing that, then people will step up and try it. And the converse applies, too, and how: if there's no perceived need for it, no one will bother. That's especially true when you're talking about making molecules that are smaller than the usual big-and-complex synthetic targets, and made via harder-than-it-looks chemistry.

Thoughts from the industrial end of things? I'd be happy to see more work like this being done, although I think it' going to take more than one paper like this to get it going. That said, the intersection with popular fragment-based drug design ideas, which are already having an effect in the purely academic world of diversity-oriented synthesis, might give an extra impetus to all this.

Comments (34) + TrackBacks (0) | Category: Chemical News | Drug Assays | Drug Development | The Scientific Literature

December 6, 2011

Riding to the Rescue of Rhodanines

Email This Entry

Posted by Derek

There's a new paper coming to the defense of rhodanines, a class of compound that has been described as "polluting the scientific literature". Industrial drug discovery people tend to look down on them, but they show up a lot, for sure.

This new paper starts off sounding like a call to arms for rhodanine fans, but when you actually read it, I don't think that there's much grounds for disagreement. (That's a phenomenon that's worth writing about sometime by itself - the disconnects between title/abstract and actual body text that occur in the scientific literature). As I see it, the people with a low opinion of rhodanines are saying "Look out! These things hit in a lot of assays, and they're very hard to develop into drugs!". And this paper, when you read the whole thing, is saying something like "Don't throw away all the rhodanines yet! They hit a lot of things, but once in a while one of them can be developed into a drug!" The argument is between people who say that elephants are big and people who say that they have trunks.

The authors prepared a good-sized assortment of rhodanines and similar heterocycles (thiohydantoins, hydantoins, thiazolidinediones) and assayed them across several enzymes. Only the ones with double-bonded sulfur (rhodanines and thiohydantoins) showed a lot of cross-enzyme potency - that group has rather unusual electronic properties, which could be a lot of the story. Here's the conclusion, which is what makes me think that we're all talking about the same thing:

We therefore think that rhodanines and related scaffolds should not be regarded as problematic or promiscuous binders per se. However, it is important to note that the intermolecular interaction profile of these scaffolds makes them prone to bind to a large number of targets with weak or moderate affinity. It may be that the observed moderate affinities of rhodanines and related compounds, e.g. in screening campaigns, has been overinterpreted in the past, and that these compounds have too easily been put forward as lead compounds for further development. We suggest that particularly strong requirements, i.e. affinity in the lower nanomolar range and proven selectivity for the target, are applied in the further assessment of rhodanines and related compounds. A generalized "condemnation" of these chemotypes, however, appears inadequate and would deprive medicinal chemists from attractive building blocks that possess a remarkably high density of intermolecular interaction points.

That's it, right there: the tendency to bind off-target, as noted by these authors, is one of the main reasons that these compounds are regarded with suspicion in the drug industry. We know that we can't test for everything, so when you have one of these structures, you're always fearful of what else it can do once it gets into an animal (or a human). Those downstream factors - stability, pharmacokinetics, toxicity - aren't even addressed in this paper, which is all about screening hits. And that's another source of the bad reputation, for industry people: too many times, people who aren't so worried about those qualities have screening commercial compound collections, come up with rhodanines, and published them as potential drug leads, when (as this paper illustrates), you have to be careful even using them as tool compounds. Given a choice, we'd just rather work on something else. . .

Comments (7) + TrackBacks (0) | Category: Drug Assays | Drug Development | The Scientific Literature

November 7, 2011

Rating A Massive Pile of Compounds

Email This Entry

Posted by Derek

Here's an interesting exercise carried out in the medicinal chemistry departments at J&J. The computational folks took all the molecules in the company's files, and then all the commercially available ones (over five million compounds), minus natural products, which were saved for another effort, and minus the obviously-nondruglike stuff (multiple nitro groups, solid hydrocarbons with no functionality, acid chlorides, etc.) They then clustered things down into (merely!) about 20,000 similarity clusters, and asked the chemists to rate them with up, down, or neutral votes.

What they found was that the opinions of the med-chem staff seemed to match known drug-like properties very closely. Molecular weights in the 300 to 400 range were most favorably received, while the likelihood of a downvote increased below 250 or above 425 or so. Similar trends held for rotatable bonds, hydrogen bond donors and acceptors, clogP, and other classic physical property descriptors. Even the ones that are hard to eyeball, like polar surface area, fell into line.

It's worth asking if that's a good thing, a bad thing, or nothing surprising at all. The authors themselves waffle a bit on that point:

The results of our experiment are fully consistent with prior literature on what confers drug- or lead-like characteristics to a chemical substance. Whether the strategy will yield the desired results in the long term with respect to quality, novelty, and number of hits/leads remains to be seen. It is also unclear whether this strategy can lead to sufficient differentiation from a competitive stand-point. In the meantime, the only undeniable benefits we can point to is that we harnessed our chemists’ opinions to select lead-like molecules that are totally within reasonable property ranges, that fill diversity holes, and that have been purchased for screening, and that we did so in a way that promoted greater transparency, greater awareness, greater collaboration, and a renewed sense of involvement and engagement of our employees.

I'll certainly give them the diversity-of-the-screening-deck point. But I'm not so sure about that renewed sense of involvement stuff. Apparently 145 chemists participated in total (this effort was open to everyone), but no mention is made of what fraction of the total staff that might be. People were advised to try to vote on at least 2,000 clusters (!), but fewer than half the participants even made it that far. Ten people made it halfway through the lot, and 6 lunatics actually voted on every single one of the 22,015 clusters, which makes me think that they had way too much time on their hands and/or have interesting and unusual personality features. A colleague's reaction to that figure was "Wow, they'll have to track those people down", to which my uncharitable reply was "Yeah, with a net".

So while this paper is interesting to read, I can't say that I would have been all that happy participating in it (although I've certainly had smaller-scale experiences of this type). And I'd like to know what the authors thought when they finally assembled all the votes and realized that they'd recapitulated a set of filters that they could have run in a few seconds, since they're surely already built into their software. And we all should reflect on how thoroughly we seem to have incorporated Lipinski's rules into our own software, between our ears. On balance, it's probably a good thing, but it's not without a price.

Comments (16) + TrackBacks (0) | Category: Drug Assays | Life in the Drug Labs

October 26, 2011

Francis Collins Speaks

Email This Entry

Posted by Derek

With all the recent talk about the NIH's translational research efforts, and the controversy about their drug screening efforts, this seems like a good time to note this interview with Francis Collins over at BioCentury TV. (It's currently the lead video, but you'll be able to find it in their "Show Guide" afterwards as well).

Collins says that they're not trying to compete with the private sector, but taking a look at the drug development process "the way an engineer would", which takes me back to this morning's post re: Andy Grove. One thing he emphasizes is that he believes that the failure rate is too high because the wrong targets are being picked, and that target validation would be a good thing to improve.

He's also beating the drum for new targets to come out of more sequencing of human genomes, but that's something I'll reserve judgment on. The second clip has some discussion of the DARPA-backed toxicology chip and some questions on repurposing existing drugs. The third clip talks about the FDA's role in all this, and tries to clarify what NIH's role would be in outlicensing any discoveries. (Collins also admits along the way that the whole NCATS proposal has needed some clarifying as well, and doesn't sound happy with some of the press coverage).

Part 5 (part 4 is just a short wrap-up) discusses the current funding environment, and then moves into ethics and conflicts of interest - other people's conflicts, I should note. Worth a lunchtime look!

Comments (16) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development

October 21, 2011

Does Anyone Want the NIH's Drug Screening Program?

Email This Entry

Posted by Derek

Science is reporting some problems with the NIH's drug screening efforts:

A $70-million-a-year program launched 7 years ago at the National Institutes of Health (NIH) to help academic researchers move into industry-style drug discovery may soon be forced to scale back sharply. NIH Director Francis Collins has been one of its biggest champions. But the NIH Molecular Libraries, according to plan, must be weaned starting next year from the NIH director's office Common Fund and find support at other NIH institutes. In a time of tight budgets, nobody wants it.

The fate of the Molecular Libraries program became “an extremely sensitive political issue” earlier this year when NIH realized it would not be easy to find a new home for the program, said one NIH official speaking on background. . .

. . .John Reed, head of the Sanford-Burnham Medical Research Institute screening center in San Diego, which receives about $16 million a year from the Common Fund, says his center has so far attracted only modest funding from drug companies. He expressed frustration with the Common Fund process. “NIH has put a huge investment into [the Molecular Libraries], and it's running very well,” he says. “If there's not a long-term commitment to keep it available to the academic community, why did we make this hundreds of millions of dollars investment?”

Good question! This all grew out of the 2003 "NIH Roadmap" initiative - here's a press release from better days. But it looks partly to be a victim of sheer bad timing. There's not a lot of extra money sloshing around the drug industry these days, and there sure isn't a lot in NIH's budget, either. You wouldn't know that there's a problem at all from looking at the program's web site, would you?

Since I know there are readers out there from both sides of this particular fence, I'd be interesting in hearing some comments. Has the screening initiative been worthwhile? Should it be kept up - and if so, how?

Comments (21) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays

October 7, 2011

Different Drug Companies Make Rather Different Compounds

Email This Entry

Posted by Derek

Now here's a paper, packed to the edges with data, on what kinds of drug candidate compounds different companies produce. The authors assembled their list via the best method available to outsiders: they looked at what compounds are exemplified in patent filings

What they find is that over the 2000-2010 period that not much change has taken place, on average, in the properties of the molecules that are showing up. Note that we're assuming, for purposes of discussion, that these properties - things like molecular weight, logP, polar surface area, amount of aromaticity - are relevant. I'd have to say that they are. They're not the end of the discussion, because there are plenty of drugs that violate one or more of these criteria. But there are even more that don't, and given the finite amount of time and money we have to work with, you're probably better off approaching a new target with five hundred thousand compounds that are well within the drug-like properties boxes rather than five hundred thousand that aren't. And at the other end of things, you're probably better off with ten clinical candidates that mostly fit versus ten that mostly don't.

But even if overall properties don't seem to be changing much, that doesn't mean that there aren't differences between companies. That's actually the main thrust of the paper: the authors compare Abbott, Amgen, AstraZeneca, Bayer-Schering, Boehringer, Bristol-Myers Squibb, GlaxoSmithKline, J&J, Lilly, Merck, Novartis, Pfizer, Roche, Sanofi, Schering-Plough, Takeda, Wyeth, and Vertex. Of course, these organizations filed different numbers of patents, on different targets, with different numbers of compounds. For the record, Merck and GSK filed the most patents during those ten years (over 1500), while Amgen and Takeda filed the fewest (under 300). Merck and BMS had the largest number of unique compounds (over 70,000), and Takeda and Bayer-Schering had the fewest (in the low 20,000s). I should note that AstraZeneca just missed the top two in both patents and compounds.
If you just look at the raw numbers, ignoring targeting and therapeutic areas, Wyeth, Bayer-Schering, and Novartis come out looking the worst for properties, while Vertex and Pfizer look the best. But what's interesting is that even after you correct for targets and the like, that organizations still differ quite a bit in the sorts of compounds that they turn out. Takeda, Lilly, and Wyeth, for example, were at the top of the cLogP rankings (numberically, "top" meaning the greasiest). Meanwhile, Vertex, Pfizer, and AstraZeneca were at the other end of the scale in cLogP. In molecular weight, Novartis, Boehringer, and Schering-Plough were at the high end (up around 475), while Vertex was at the low end (around 425). I'm showing a radar-style plot from the paper where they cover several different target-unbiased properties (which have been normalized for scale), and you can see that different companies do cover very different sorts of space. (The numbers next to the company names are the total number of shared targets found and the total number of shared-target observations used - see the paper if you need more details on how they compiled the numbers).

Now, it's fair to ask how relevant the whole sweep of patented compounds might be, since only a few ever make it deep into the clinic. And some companies just have different IP approaches, patenting more broadly or narrowly. But there's an interesting comparison near the end of the paper, where the authors take a look at the set of patents that cover only single compounds. Now, those are things that someone has truly found interesting and worth extra layers of IP protection, and they average to significantly lower molecular weights, cLogP values, and number of rotatable bonds than the general run of patented compounds. Which just gets back to the points I was making in the first paragraph - other things being equal, that's where you'd want to spend more of your time and money.

What's odd is that the trends over the last ten years haven't been more pronounced. As the paper puts it:

blockquote>Over the past decade, the mean overall physico-chemical space used by many pharmaceutical companies has not changed substantially, and the overall output remains worryingly at the periphery of historical oral drug chemical space. This is despite the fact that potential candidate drugs, identified in patents protecting single compounds, seem to reflect physiological and developmental pressures, as they have improved drug-like properties relative to the full industry patent portfolio. Given these facts, and the established influence of molecular properties on ADMET risks and pipeline progression, it remains surprising that many organizations are not adjusting their strategies.

The big question that this paper leaves unanswered, because there's no way for them to answer it, is how these inter-organizational differences get going and how they continue. I'll add my speculations in another post - but speculations they will be.

Comments (30) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

September 14, 2011

Lilly's Open Screening Program: An Update

Email This Entry

Posted by Derek

I'm still at the RSC/SCI symposium in Cambridge, and a talk yesterday by Marta Pineiro-Nuñez gives me a chance to update this post about Eli Lilly's foray into opening up its screening to outside collaboration. That effort has been working away for the last two or three years, and the company is now revealing some details about how it's been going.

The original plan was to allow people to put compounds into a set of Lilly phenotypic screens. No structures would be revealed, and the company would have "first rights of negotiated access" for any interesting hits. They have a new web gateway for the whole thing, since now they've added several target-based screens to the process. As mentioned in the earlier post, they've come up with a universal Material Transfer Agreement to bring the compounds in, but Pineiro-Nuñez said that this was still a bit of a struggle at first. Small companies were pretty open to the idea, she said, but there were some suspicious responses from academia, with a lot of careful digging through the MTA to make sure that they wouldn't be giving away too much.

But things seem to have gotten going pretty well. According to the presentation, Lilly has 252 affiliations in 27 countries. That breaks down as 174 academic partners and 78 small companies. About 42,000 compounds have been accepted for screening - that's after a firewalled computational screen of the structures to eliminate nasty functional groups and the like. About 40% of the submissions fail the suitability screens, but the single biggest reason is lack of structural novelty - too close to marketed drugs, too close to controlled substances, or too close to things that are already in Lilly's files.

Here's a recent overview of the screening results. In the end, 115 structures were requested for disclosure, and 97 of those ended up being shared with Lilly, who still wanted 13 of them after looking them over. And those have (so far) led to two recent signed collaborations, with one more set to go and two others still in negotiations. The compounds certainly aren't instant clinical candidates, but have been interesting enough to put money on. And so far, the initiative is seen as successful, enough to expand it to more assays.

It'll be interesting to see if more companies try this out. It would seem especially suited for unusual proprietary assays that might be hiding behind industrial walls. Having Lilly demonstrate that a model of this sort can actually work in practice should help - congratulations to them for putting the work in to make it happen.

Comments (10) + TrackBacks (0) | Category: Drug Assays

September 7, 2011

Get Yer Rhodanines Here

Email This Entry

Posted by Derek

We've talked here before about the structural class known as rhodanines - the phrase "polluting the scientific literature" has been used to describe them, since they rather promiscuously light up a lot of drug target assays, and almost never to any useful effect.

Well, guess what? Now there's an even easier way to make them! And says this new paper in the Journal of Organic Chemistry:

5-(Z)-Alkylidene-2-thioxo-1,3-thiazolidin-4-ones (rhodanine derivatives) were prepared by reaction of in situ generated dithiocarbamates with recently reported racemic α-chloro-β,γ-alkenoate esters. This multicomponent sequential transformation performed in one reaction flask represents a general route to this medicinally valuable class of sulfur/nitrogen heterocycles. Using this convergent procedure, we prepared an analogue of the drug epalrestat, an aldose reductase inhibitory rhodanine.
Sequentially linking several different components in one reaction vessel has been studied intensively as a rapid way to increase molecular complexity while avoiding costly and environmentally unfriendly isolation and purification of intermediates.(1-4) Such efficient multicomponent reactions, such as the Ugi reaction, often produce privileged scaffolds of considerable medicinal value. Rhodanines (2-thioxo-1,3-thiazolidin-4-ones) are five-membered ring sulfur/nitrogen heterocycles some of which have antimalarial, antibacterial, antifungal, antiviral, antitumor, anti-inflammatory, or herbicidal activities. . .In conclusion, convergent syntheses of N-alkyl 5-(Z)-alkylidene rhodanine derivatives have been achieved using recently reported racemic α-chloro-β,γ-alkenoate ester building blocks. The formation of these rhodanine derivatives involves a three-step, one-flask protocol that provides quick access to biologically valuable sulfur–nitrogen heterocycles.

Just what we needed. Now it's only going to be a matter of time before someone makes and sells a library of these things, and we can all get to see them again as screening hits in the literature.

Comments (11) + TrackBacks (0) | Category: Chemical News | Drug Assays

September 2, 2011

How Many New Drug Targets Aren't Even Real?

Email This Entry

Posted by Derek

So, are half the interesting new results in the medical/biology/med-chem literature impossible to reproduce? I linked earlier this year to an informal estimate from venture capitalist Bruce Booth, who said that this was his (and others') experience in the business. Now comes a new study from Bayer Pharmaceuticals that helps put some backing behind those numbers.

To mitigate some of the risks of such investments ultimately being wasted, most pharmaceutical companies run in-house target validation programmes. However, validation projects that were started in our company based on exciting published data have often resulted in disillusionment when key data could not be reproduced. Talking to scientists, both in academia and in industry, there seems to be a general impression that many results that are published are hard to reproduce. However, there is an imbalance between this apparently widespread impression and its public recognition. . .

Yes, indeed. The authors looked back at the last four years worth of oncology, women's health, and cardiovascular target validation efforts inside Bayer (this would put it right after they combined with Schering AG of Berlin). They surveyed all the scientists involved in early drug discovery in those areas, and had them tally up the literature results they'd acted on and whether they'd panned out or not. I should note that this is the perfect place to generate such numbers, since the industry scientists are not in it for publication glory, grant applications, or tenure reviews: they're interested in finding drug targets that look like they can be prosecuted, in order to find drugs that could make them money. You may or may not find those to be pure or admirable motives (I have no problem at all with them, personally!), but I think we can all agree that they're direct and understandable ones. And they may be a bit orthogonal to the motives that led to the initial publications. . .so, are they? The results:

"We received input from 23 scientists (heads of laboratories) and collected data from 67 projects, most of them (47) from the field of oncology. This analysis revealed that only in ~20–25% of the projects were the relevant published data completely in line with our in-house findings. In almost two-thirds of the projects, there were inconsistencies between published data and in-house data that either considerably prolonged the duration of the target validation process or, in most cases, resulted in termination of the projects. . ."

So Booth's estimate may actually have been too generous. How does this gap get so wide? The authors suggest a number of plausible reasons: small sample sizes in the original papers, leading to statistical problems, for one. The pressure to publish in academia has to be a huge part of the problem - you get something good, something hot, and you write that stuff up for the best journal you can get it into - right? And it's really only the positive results that you hear about in the literature in general, which can extend so far as (consciously or unconsciously) publishing just on the parts that worked. Or looked like they worked.

But the Bayer team is not alleging fraud - just irreproducibility. And it seems clear that irreproducibility is a bigger problem than a lot of people realize. But that's the way that science works, or is supposed to. When you see some neat new result, your first thought should be "I wonder if that's true?" You may have no particular reason to doubt it, but in an area with as many potential problems as discovery of new drug targets, you don't need any particular reasons. Not all this stuff is real. You have to make every new idea perform the same tricks in front of your own audience, on your own stage under bright lights, before you get too excited.

Comments (51) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development

August 26, 2011

Kibdelomycin, A New Antibiotic. In A Way.

Email This Entry

Posted by Derek

We're going to need new antibiotics. Everyone knows this, and it's not like no one's been trying to do anything about it, either, but. . .we're still going to need more of them than we have. I'm not predicting that we're going to go all the way back to a world where young, healthy people with access to the best medical care die because they decided to play tennis without their socks on, but we're certainly in danger of a much nastier world than we have.

So I'm always interested to hear of new antibiotic discovery programs, and Merck is out with an interesting paper on theirs. They've been digging through the natural products, which have been the fount from which almost all antibiotics have sprung, and they have a new one called kibdelomycin to report. This one was dug out from an organism in a sample from the Central African Republic by a complicated but useful screening protocol, the S. aureus fitness test. This relies on 245 different engineered strains of the bacterium, each with an inducible RNAi pathway to downregulate some essential gene. When you pool these into mixed groups and grow them in the presence of test compounds (or natural product extracts) for 20 generations or so, a check of what strains have moved ahead (and fallen behind) can tell you what pathways you seem to be targeting. A key feature is that you can compare the profile you get with those of known antibiotics, so you don't end up rediscovering something (or discovering something that only duplicates what we already have anyway).
Kibdelomycin.png Now, that's no one's idea of a beautiful structure, although (to be fair) a lot of antibiotics have very weird structures themselves. But it's safe to say that there are some features there that could be trouble in a whole animal, such as that central keto-enol-pyrrolidone ring and the funky unsaturated system next to it. (The dichloropyrrole, though, is interestingly reminiscent of these AstraZeneca gyrase/topoisomerase antibiotic candidates, while both celestramycin and pyoluteorin have a different dichloropyrrole in them).

What kind of activity does kibdelomycin have? Well, this is where my enthusiasm cools off just a bit more. It showed up in screening with a profile similar to the coumarin antibiotics novobiocin and chlorobiocin, and sure enough, it's a topoisomerase II inhibitor. It appears to be active almost entirely on gram-positive organisms. And while there are certainly nasty gram-positive infections that have to be dealt with, I'm more encouraged when I see something that hits gram-negatives as well. They've got more complicated defenses, those guys, and they're harder to kill. It's not easy to get broad-spectrum activity when you're going after gyrase/Topo II, but the fluoroquinolones definitely manage it.

The Merck team makes much out of kibdelomycin being "the first truly novel bacterial type II topoisomerase inhibitor with potent antibacterial activity discovered from natural product sources in more than six decades". And they're right that this is an accomplishment. But the kicker in that sentence is "from natural product sources". Getting gram-positive Topo II inhibitors has actually been one of the areas where synthetic compounds have had the most success. Building off the quinolones themselves has been a reasonably fruitful strategy, and a look through the literature turns up a number of other structural classes with this sort of activity (including some pretty wild ones). Not all of these are going places, but there are certainly a number of possibilities out there.

In short, if kibdelomycin weren't an odd-looking natural product, I wonder how much attention another high-molecular-weight gram-positive-only topoisomerase inhibitor would be getting, especially with only in vitro data behind it. Every little bit helps, and having a new structural class to work from is a worthwhile discovery. But one could still want (and hope) for more.

Comments (14) + TrackBacks (0) | Category: Drug Assays | Infectious Diseases | Natural Products

August 18, 2011

Is Anyone Doing the Pfizer Screening Deal?

Email This Entry

Posted by Derek

A couple of years ago, I wrote here about an initiative from Pfizer. They were proposing letting other (smaller) companies screen their compound collection, with rights to be worked out if something interesting turned up.

The thing is, I haven't heard about anyone taking them up on it. Does anyone know if this ever got off the ground, or did it get lost in the trackless Pfizer territories somewhere? It sounded like a reasonable idea in some ways, and I'm curious if it ever went anywhere. . .

Comments (13) + TrackBacks (0) | Category: Drug Assays

July 27, 2011

Bait And Switch For Type B GPCRs

Email This Entry

Posted by Derek

You hear often about how many marketed drugs target G-protein coupled receptors (GPCRs). And it's true, but not all GPCRs are created equal. There's a family of them (the Class B receptors) that has a number of important drug targets in it, but getting small-molecule drugs to hit them has been a real chore. There's Glucagon, CRF, GHRH, GLP-1, PACAP and plenty more, but they all recognize good-sized peptides as ligands, not friendly little small molecules. Drug-sized things have been found that affect a few of these receptors, but it has not been easy, and pretty much all of them have been antagonists. (That makes sense, because it's almost always easier to block some binding event rather than hitting the switch just the right way to turn a receptor on).

That peptide-to-receptor binding also means that we don't know nearly as much about what's going on in the receptor as we do for the small-molecule GPCRs, either (and there are still plenty of mysteries around even those). The generally accepted model is a two-step process: there's an extra section of the receptor protein that sticks out and recognizes the C-terminal end of the peptide ligand first. Once that's bound, the N-terminal part of the peptide ligand binds into the seven-transmembrane-domain part of the receptor. The first part of that process is a lot more well-worked-out than the second.

Now a German team has reported an interesting approach that might help to clear some things up. They synthesized a C-terminal peptide that was expected to bind to the extracellular domain of the CRF receptor, and made it with an azide coming off its N-terminal end. (Many of you will now have guessed where this is going!) Then they took a weak peptide agonist piece and decorated its end with an acetylene. Doing the triazole-forming "click" reaction between the two gave a nanomolar agonist for the receptor, revving up the activity of the second peptide by at least 10,000x.

This confirms the general feeling that the middle parts of the peptide ligands in this class are just spacers to hold the two business ends together in the right places. But it's a lot easier to run the "click" reaction than it is to make long peptides, so you can mix and match pieces more quickly. That's what this group did next, settling on a 12-amino-acid sequence as their starting point for the agonist peptide and running variations on it.

Out of 89 successful couplings to the carrier protein, 70 of the new combinations lowered the activity (or got rid of it completely). 15 were about the same as the original sequence, but 11 of them were actually more potent. Combining those single-point changes into "greatest-hit" sequences led to some really potent compounds, down to picomolar levels. And by that time, they found that they could get rid of the tethered carrier protein part, ending up with a nanomolar agonist peptide that only does the GPCR-binding part and bypasses the extracellular domain completely. (Interestingly, this one had five non-natural amino acid substitutions).

Now that's a surprise. Part of the generally accepted model for binding had the receptor changing shape during that first extracellular binding event, but in the case of these new peptides, that's clearly not happening. These things are acting more like the small-molecule GPCR agonists and just going directly into the receptor to do their thing. The authors suggest that this "carrier-conjugate" approach should speed up screening of new ligands for the other receptors in this category, and should be adaptable to molecules that aren't peptides at all. That would be quite interesting indeed: leave the carrier on until you have enough potency to get rid of it.

Comments (3) + TrackBacks (0) | Category: Biological News | Chemical News | Drug Assays

July 7, 2011

Phenotypic Screening For the Win

Email This Entry

Posted by Derek

Here's another new article in Nature Reviews Drug Discovery that (for once) isn't titled something like "The Productivity Crisis in Drug Research: Hire Us And We'll Consult Your Problems Away". This one is a look back at where drugs have come from.

Looking over drug approvals (259 of them) between 1999 and 2008, the authors find that phenotypic screens account for a surprising number of the winners. (For those not in the business, a phenotypic screen is one where you give compounds to some cell- or animal-based assay and look for effects. That's in contrast to the target-based approach, where you identify some sort of target as being likely important in a given disease state and set out to find a molecule to affect it. Phenotypic screens were the only kinds around in the old days (before, say, the mid-1970s or thereabouts), but they've been making a comeback - see below!)

Out of the 259 approvals, there were 75 first-in-class drugs and 164 followers (the rest were imaging agents and the like). 100 of the total were discovered using target-based approaches, 58 through phenotypic approaches, and 18 through modifying natural substances. There were also 56 biologics, which were all assigned to the target-based category. But out of the first-in-class small molecules, 28 of them could be assigned to phenotypic assays and only 17 to target-based approaches. Considering how strongly tilted the industry has been toward target-based drug discovery, that's really disproportionate. CNS and infectious disease were the therapeutic areas that benefited the most from phenotypic screening, which makes sense. We really don't understand the targets and mechanisms in the former, and the latter provide what are probably the most straightforward and meaningful phenotypic assays in the whole business. The authors' conclusion:

(this) leads us to propose that a focus on target-based drug discovery, without accounting sufficiently for the MMOA (molecular mechanism of action) of small-molecule first-in-class medicines, could be a technical reason contributing to high attrition rates. Our reasoning for this proposal is that the MMOA is a key factor for the success of all approaches, but is addressed in different ways and at different points in the various approaches. . .

. . .The increased reliance on hypothesis-driven target-based approaches in drug discovery has coincided with the sequencing of the human genome and an apparent belief by some that every target can provide the basis for a drug. As such, research across the pharmaceutical industry as well as academic institutions has increasingly focused on targets, arguably at the expense of the development of preclinical assays that translate more effectively into clinical effects in patients with a specific disease.

I have to say, I agree (and have said so here on the blog before). It's good to see some numbers put to that belief, though. This, in fact, was the reason why I thought that the NIH funding for translational research might be partly spent on new phenotypic approaches. Will we look back on the late 20th century/early 21st as a target-based detour in drug discovery?

Comments (36) + TrackBacks (0) | Category: Drug Assays | Drug Development | Drug Industry History

May 6, 2011

In Which I Do Not Lose It, For Once

Email This Entry

Posted by Derek

PNAS recently came out with a special concentration of chemistry papers, and they're worth a look. The theme is the synthesis of chemical probes, which makes me think that maybe Stuart Schrieber can guest-edit an issue of Vogue next. Today I'm going to highlight one from the Broad Institute on diversity-oriented synthesis (DOS), and next week I'll get to some more.

OK, that was something of a come-on for regular readers of this site, who now will be listening for the sound of grinding wheels coming up to speed, the better to sharpen the Sword of Justice. I've said unfriendly things in the past about DOS and some of the claims made for it. The point of much of this work has been lost on me, and I'm a pretty broad-minded guy. (That word, in this case, rhymes with "sawed", not with "load"). The first flush (no aspersions meant) of papers in the field might just as well have been titled "Check It Out: A Bunch of Huge Compounds No One's Ever Made Before", and were followed up, in my mind, by landmark publications such as "A Raving Heapload of Structures You Didn't Want in the First Place" and "Dang, There Are Even More Compounds With Molecular Weight 850 Than We Thought". But does it have to be this way?

Maybe not. As I mentioned earlier this year, people are starting to compare DOS and fragment-based approaches. (I think that Nature dialog could have been more useful than it was, but it was a start). And this latest paper continues that process. It's using DOS approaches to generate smaller molecular weight compounds - fragments, actually. They're not tiny ones, more medium-to-large size by fragment-based standards, but they're under 300 MW.

And, importantly, they're deliberately designed to be three-dimensional - lots of pyrrolidines and fused-ring compounds thereof, homopiperidines, spiro-lactams, and so on. Many of the early fragment libraries (and many of the commercial ones that you can still buy) are too invested in small, flat, heterocycles. It's not that you can't get good leads from those things, but there's a lot more to life (and to molecular property space). This paper's collection is still a bit heavy on the alkenes to my taste (all those ring-closing metathesis reactions), but they've also reduced those for part of the library, which means that a screen of this collection will tell you if the olefin is a key structural feature or not. The alkenes themselves could serve as useful handles to build out from as well; a fragment hit with no ways to elaborate its structure isn't too useful.

As I said back in February, "I'd prefer that DOS collections not get quite so carried away, and explore new structural motifs more in the range of druglike space." That's exactly what this paper does, and I think its direction should be encouraged. This plays to the strengths of both approaches, rather than pushing either of them to the point where they break down.

Comments (12) + TrackBacks (0) | Category: Chemical News | Drug Assays

May 3, 2011

A Look Inside the Compound Collections

Email This Entry

Posted by Derek

Now here's a comparison that you don't get to see very often: how much do two large pharma compound collections overlap? There's a paper going into just that question in the wake of the 2006-2007 merger between Bayer and Schering AG. (By two coincidences, this paper is in the same feed as the one that I highlighted yesterday, and that merger is the one that closed my former research site out from under me).

Pre-merger, Bayer had over two million structures in its corporate collection, and Schering AG had just under 900,000. Both companies had undertaken recent library clean-up programs, clearing out undesirable compounds and adding both purchased and in-house diversity structures. Interestingly, it turns out that just under 50,000 structures were duplicated across both collections, about 1.5% of the total. Almost all of these duplicates were purchased compounds; only 2,000 of them had been synthesized in-house. And even most of those turned out to be from combichem programs or were synthetic intermediates - there was almost no overlap at all in submitted med-chem compounds.

Various measures of structural complexity and similarity backed up those numbers. The two collections were surprisingly different, which might well have something to do with the different therapeutic areas the two companies had focused on over the years. The Bayer compounds tended to run higher in molecular weight, rotatable bonds, and clogP, but then, a higher percentage of the Schering AG compounds were purchased with such filters already in place. As for undesirable structures, only about 2% of the Bayer collection and 1% of the Schering AG compounds were considered to be real offenders. I hope none of those were mine; I contributed quite a few compounds to the collection over the years, but they were, for the most part, relatively sane.

The paper's conclusion can be read in more than one way:

Furthermore, an argument that might support mergers and acquisitions (M&A) in the pharmaceutical sector can be harvested from this analysis. Currently, M&As in this industry are driven by product portfolios rather than by drug discovery competencies. With the current need for innovative drugs, R&D skills of pharmaceutical companies might again become more important. The technological complementarity of two companies is often quoted as an important factor for successful M&As in the long term. If compound libraries are regarded as a kind of company knowledge-base, then a high degree of complementarity is clearly desirable and would improve drug discovery skills. Based on our data, the libraries of BHC and SAG are structurally complementary and fit together well in terms of their physico-chemical properties. However, it remains to be proven if this leads to additional innovative products.

Not so sure about that, myself. I don't know how good a proxy the compound collections are, since the represent an historical record as much as they do the current state of a company. And that paragraph glosses over the effect of mergers on R&D itself - it's not like just adding pieces together, that's for sure. The track record for mergers generating "additional innovative products" is not good. We'll see how the Bayer-Schering one holds up. . .

Comments (13) + TrackBacks (0) | Category: Business and Markets | Drug Assays | Drug Industry History

February 7, 2011

Fragments Versus DOS: A Showdown

Email This Entry

Posted by Derek

Nature has side-by-side editorial pieces about fragment-based drug discovery versus diversity-oriented synthesis (DOS). I've written about both topics here before (DOS here and here, fragments here and here), and it should be fairly clear that I favor the former. But both ideas deserve a hearing.

Background, for those who aren't having to think about this stuff: the fragment-based approach is to screen a reasonable set (hundreds to low thousands) of small (MW 150 to 300) molecules. You won't find any nanomolar hits that way, but you will find things that (for that molecular weight) are binding extremely efficiently. If you can get a structure (X-ray, most of the time), you can then use that piece as a starting point and build out, trying to keep the binding efficiency high as you go. Diversity-oriented synthesis, on the other hand, tries to make larger molecules that are in structural spaces not found in nature (or in other screening collections, either). It's a deliberate attempt to make wild-blue-yonder compounds in untried areas, and is often used to screen against similarly untried targets that haven't shown much in conventional screening.

The two articles make their cases, but spend some time talking past each other. Abbott's Phil Hajduk takes the following shots at DOS: that it's tended to produce compounds whose molecular weights are too high (and whose other properties are also undesirable), and that it needs (in order to cover any meaningful amount of chemical space at those molecular weights) to produce millions of compounds, all of which must then be screened. Meanwhile, Warren Galloway and David Spring of Cambridge make the following charges about fragment work: that it only works when you have a specific molecular target in mind (and that only then when you have high-quality structural information), that it tends to perform poorly against the less tractable targets (such as protein-protein interactions), and that fragments (and the molecules derived from them) tend not to be three-dimensional enough.

Here's my take: I like phenotypic screening, where you run compound collections across cells/tissues/small animal models and see what works. And fragment are indeed next to useless for that purpose. But I agree with Hajduk that most of the DOS compound libraries I've seen are far too large and ugly to furnish anything more than a new probe compound from such screens. There are many academic labs for whom that's a perfectly good end point, and they publish a paper saying, in short, We Found the First Compound That Makes X Cells Do Y. Which is interesting, and can even be important, but there's often no path whatsoever from that compound to an actual drug. I'd prefer that DOS collections not get quite so carried away, and explore new structural motifs more in the range of druglike space. But that's not easy - new structures are a lot easier to come by if you're willing to make compounds with molecular weights of 500 to 1000, since (a) not so many people have made such beasts before, and (b) there are a lot more possible structures up there.

Now, if I have a defined target, and can get structures, I'd much prefer to do things the fragment way. But this is where the two editorial talk past each other - they both beat the drum for what they do well, but they do different things well. It's the parts where they overlap that I find most interesting. One of those is, as just mentioned, the problem that DOS compounds tend to be too large and undevelopable (with one solution being to go back and make them more tractable to start with). The other overlap is whether fragment collections can hit well against tough targets like protein-protein interactions. I don't know the answer to that one myself - I'd be glad to hear of examples both pro and con.

So we'll call this a struggle still in progress. With any luck, both techniques will keep each other's partisans on their toes and force them to keep improving.

Comments (33) + TrackBacks (0) | Category: Drug Assays | Drug Development

February 1, 2011

The NIH's New Drug Discovery Center: Heading Into the Swamp?

Email This Entry

Posted by Derek

I've been meaning to comment on the NIH's new venture into drug discovery, the National Center for Advancing Translational Sciences. Curious Wavefunction already has some thoughts here, and I share his concerns. We're both worried about the gene-o-centric views of Francis Collins, for example:

Creating the center is a signature effort of Dr. Collins, who once directed the agency’s Human Genome Project. Dr. Collins has been predicting for years that gene sequencing will lead to a vast array of new treatments, but years of effort and tens of billions of dollars in financing by drug makers in gene-related research has largely been a bust.

As a result, industry has become far less willing to follow the latest genetic advances with expensive clinical trials. Rather than wait longer, Dr. Collins has decided that the government can start the work itself.

“I am a little frustrated to see how many of the discoveries that do look as though they have therapeutic implications are waiting for the pharmaceutical industry to follow through with them,” he said.

Odd how the loss of tens of billions of dollars - and vast heaps of opportunity cost along the way - will make people reluctant to keep going. And where does this new center want to focus in particular? The black box that is the central nervous system:

Both the need for and the risks of this strategy are clear in mental health. There have been only two major drug discoveries in the field in the past century; lithium for the treatment of bipolar disorder in 1949 and Thorazine for the treatment of psychosis in 1950.

Both discoveries were utter strokes of luck, and almost every major psychiatric drug introduced since has resulted from small changes to Thorazine. Scientists still do not know why any of these drugs actually work, and hundreds of genes have been shown to play roles in mental illness — far too many for focused efforts. So many drug makers have dropped out of the field.

So if there are far too many genes for focused efforts (a sentiment with which I agree), what, exactly, is this new work going to focus on? Wavefunction, for his part, suggests not spending so much time on the genetic side of things and working, for example, on one specific problem, such as Why Does Lithium Work for Depression? Figuring that out in detail would have to tell us a lot about the brain along the way, and boy, is there a lot to learn.

Meanwhile, Pharmalot links to a statement from the industry trade group (PhRMA) which is remarkably vapid. It boils down to "research heap good", while beating the drum a bit for the industry's own efforts. And as an industrial researcher myself, it would be easy for me to continue heaping scorn on the whole NIH-does-drug-discovery idea.

But I actually wish them well. There really are a tremendous number of important things that we don't know about this business, and the more people working on them, the better. You'd think. What worries me, though, is that I can't help but believe that a good amount of the work that's going to be done at this new center will be misapplied. I'm really not so sure that the gene-to-disease-target paradigm just needs more time and money thrown at it, for example. And although there will be some ex-industry people around, the details of drug discovery are still likely to come as a shock to the more academically oriented people.

Put simply, the sorts of discoveries and project that make stellar academic careers, that get into Science and Nature and all the rest of them, are still nowhere near what you need to make an actual drug. It's an odd combination of inventiveness and sheer grunt work, and not everyone's ready for it. One likely result is that some people will just avoid the stuff as much as possible and spend their time and money doing something else that pleases them more.

What do I think that they should be doing, then? One possibility is the Pick One Big Problem option that Wavefunction suggests. What I'd recommend would also go against the genetic tracery stuff: I'd put money into developing new phenotypic assays in cells, tissues, and whole animals. Instead of chasing into finer and finer biochemical details in search of individual targets, I'd try to make the most realistic testbeds of disease states possible, and let the screening rip on that. Targets can be chased down once something works.

But it doesn't sound like that's what's going to happen. So, reluctantly, I'll make a prediction: if years of effort and billions of dollars thrown after genetic target-based drug discovery hasn't worked out, when done by people strongly motivated to make money off their work, then an NIH center focused on the same stuff will, in all likelihood, add very little more. It's not like they won't stay busy. That sort of work can soak up all the time and money that you can throw at it. And it will.

Comments (36) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development | Drug Industry History

November 11, 2010

And One Was Just Right?

Email This Entry

Posted by Derek

I've been reading an interesting new paper from Stuart Schreiber's research group(s) in PNAS. But I'm not sure if the authors and I would agree on the reasons that it's interesting.

This is another in the series that Schreiber has been writing on high-throughput screening and diversity-oriented synthesis (DOS). As mentioned here before, I have trouble getting my head around the whole DOS concept, so perhaps that's the root of my problems with this latest paper. In many ways, it's a companion to one that was published earlier this year in JACS. In that paper, he made the case that natural products aren't quite the right fit for drug screening, which fit with an earlier paper that made a similar claim for small-molecule collections. Natural products, the JACS paper said, were too optimized by evolution to hit targets that we don't want, while small molecules are too simple to hit a lot of the targets that we do. Now comes the latest pitch.

In this PNAS paper, Schreiber's crew takes three compound collections: 6,152 small commercial molecules, 2,477 natural products, and 6,623 from academic synthetic chemistry (with a preponderance of DOS compounds), for a total of 15, 252. They run all of these past a set of 100 proteins using their small-molecule microarray screening method, and look for trends in coverage and specificity. What they found, after getting rid of various artifacts, was that about 3400 compounds hit at least one protein (and if you're screening 100 proteins, that's a perfectly reasonable result). But, naturally, these hits weren't distributed evenly among the three compound collections. 26% of the academic compounds were hits, and 23% of the commercial set, but only 13% of the natural products.

Looking at specificity, it appears that the commercial compounds were more likely, when they hit, to hit six or more different proteins in the set, and the natural products the least. Looking at it in terms of compounds that hit only one or two targets gave a similar distribution - in each case, the DOS compounds were intermediate, and that turns out to be a theme of the whole paper. They analyzed the three compound collections for structural features, specifically their stereochemical complexity (chiral carbons as a per cent of all carbons) and shape complexity (sp3 carbons as a percent of the whole). And that showed that the commercial set was biased towards the flat, achiral side of things, while the natural products were the other way around, tilted toward the complex, multiple-chiral-center end. The DOS-centric screening set was right in the middle.

The take-home, then, is similar to the other papers mentioned above: small molecule collections are inadequate, natural product collections are inadequate: therefore, you need diversity-oriented synthesis compounds, which are just right. I'll let Schreiber sum up his own case:

. . .Both protein-binding frequencies and selectivities are increased among compounds having: (i) increased content of sp3-hybridized atoms relative to commercial compounds, and (ii) intermediate frequency of stereogenic elements relative to commercial (low frequency) and natural (high frequency) compounds. Encouragingly, these favorable structural features are increasingly accessible using modern advances in the methods of organic synthesis and commonly targeted by academic organic chemists as judged by the compounds used in this study that were contributed by members of this community. On the other hand, these features are notably deficient in members of compound collections currently widely used in probe- and drug-discovery efforts.

But something struck me while reading all this. The two metrics used to characterize these compound collections are fine, but they're also two that would be expected to distinguish them thoroughly - after all, natural products do indeed have a lot of chiral carbons, and run-of-the-mill commercial screening sets do indeed have a lot of aryl rings in them. There were several other properties that weren't mentioned at all, so I downloaded the compound set from the paper's supporting information and ran it through some in-house software that we use to break down such things.

I can't imagine, for example, evaluating a compound collection without taking a look at the molecular weights. Here's that graph - the X axis is the compound number, Y-axis is weight in Daltons:
The three different collections show up very well this way, too. The commercial compounds (almost every one under 500 MW) are on the left. Then you have that break of natural products in the middle, with some real whoppers. And after that, you have the various DOS libraries, which were apparently entered in batches, which makes things convenient.

Notice, for example that block of them standing up around 15,000 - that turns out to be the compounds from this 2004 Schreiber paper, which are a bunch of gigantic spirooxindole derivatives. In this paper, they found that this particular set was an outlier in the academic collection, with a lot more binding promiscuity than the rest of the set (and they went so far as to analyze the set with and without it included). The earlier paper, though, makes the case for these compounds as new probes of cellular pathways, but if they hit across so many proteins at the same time, you have to wonder how such assays can be interpreted. The experiments behind these two papers seem to have been run in the wrong order.

Note, also, that the commercial set includes a lot of small compounds, even many below 250 MW. This is down in the fragment screening range, for sure, and the whole point of looking at compounds of that molecular weight is that you'll always find something that binds to some degree. Downgrading the commercial set for promiscuous binding when you set the cutoffs that low isn't a fair complaint, especially when you consider that the DOS compounds have a much lower proportion of compounds in that range. Run a commercial/natural product/DOS comparison controlled for molecular weight, and we can talk.

I also can't imagine looking over a collection and not checking logP, but that's not in the paper, either. But here you are:
In this case, the natural products (around compound ID 7500) are much less obvious, but you can certainly see the different chemical classes standing out in the DOS set. Note, though, that those compounds explore high-logP regions that the other sets don't really touch.

How about polar surface area? Now the natural products really show their true character - looking over the structures, that's because there are an awful lot of polysaccharide-containing things in there, which will run your PSA up faster than anything:
And again, you can see the different libraries in the DOS set very clearly.

So there are a lot of other ways to distinguish these compounds, ways that (to be frank) are probably much more relevant to their biological activity. Just the molecular-weight one is a deal-breaker for me, I'm afraid. And that's before I start looking at the structures in the three collections at all. Now, that's another story.

I have to say, from my own biased viewpoint, I wouldn't pay money for any of the three collections. The natural product one, as mentioned, goes too high in molecular weight and is too polar for my tastes. I'd consider it for antibiotic drug discovery, but with gritted teeth. The commercial set can't make up its mind if it's a fragment collection or not. There are a bunch of compounds that are too small even for my tastes in fragments - 4-methylpyridine, for example. And there are a lot of ugly functional groups: imines of beta-napthylamine, which should not even get near the front door (unstable fluorescent compounds that break down to a known carcinogen? Return to sender). There are hydroxylamines, peroxides, thioureas, all kinds of things that I would just rather not spend my time on.

And what of the DOS collection? Well, to be fair, not all of it is DOS - there are a few compounds in there that I can't figure out, like isoquinoline, which you can buy from the catalog. But the great majority are indeed diversity-oriented, and (to my mind), diversity-oriented to a fault. The spirooxindole library is probably the worst - you should see the number of aryl rings decorating some of those things; it's like a fever dream - but they're not the only offenders in the "Let's just hang as many big things as we can off this sucker" category. Now, there are some interesting and reasonable DOS compounds in there, too, but there are also more endoperoxides and such. (And yes, I know that there are drug structures with endoperoxides in them, but damned few of them, and art is long while life is short). So no, I wouldn't have bought this set for screening, either; I'd have cherry-picked about 15 or 20% of it.

Summary of this long-winded post? I hate to say it, but I think this paper has its thumb on the scale. I'm just around the corner from the Broad Institute, though, so maybe a rock will come through my window this afternoon. . .

Comments (36) + TrackBacks (0) | Category: Academia (vs. Industry) | Drug Assays | Drug Development | Natural Products

October 26, 2010

Enthalpy and Entropy Again

Email This Entry

Posted by Derek

Earlier this year, I wrote here about using calorimetry in drug discovery. Years ago, people would have given you the raised eyebrow if you'd suggested that, but it's gradually becoming more popular, especially among people doing fragment-based drug discovery. After all, the binding energy that we depend on for our drug candidates is a thermodynamic property, and you can detect the heat being given off when the molecules bind well. Calorimetry also lets you break that binding energy down into its enthalpic (delta-H) and entropic (T delta-S) components, which is hard to do by other means.

And there's where the arguing starts. As I mentioned back in March, one idea that's been floating around is that better drug molecules tend to have more of an enthalpic contribution to their binding. Very roughly speaking, enthalpic interactions are often what med-chemists call "positive" ones like forming a new hydrogen bond or pi-stack, whereas entropic interactions are often just due to pushing water molecules off the protein with some greasy part of your molecule. (Note: there are several tricky double-back-around exceptions to both of those mental models. Thermodynamics is a resourceful field!) But in that way, it makes sense that more robust compounds with better properties might well be more enthalpically-driven in their binding.

But we do not live in a world bounded by what makes intuitive sense. Some people think that the examples given in the literature for this effect are the only decent examples that anyone has. At the fragment conference I attended the other week, though, a speaker from Astex (a company that's certainly run a lot of fragment optimization projects) said that they're basically not seeing it. In their hands, some lead series are enthalpy-driven as they get better, some are entropy-driven, and some switch gears as the SAR evolves. Another speaker said that they, on the other hand, do tend to go with the enthalpy-driven compounds, but I'm not sure if that's just because they don't have as much data as the Astex people do.

So as far as I'm concerned, the whole concept that I talked about in March is still in the "interesting but unproven" category. We're all looking for new ways to pick better starting compounds or optimize leads, but I'm still not sure if this is going to do the trick. . .

Comments (18) + TrackBacks (0) | Category: Analytical Chemistry | Drug Assays | Life in the Drug Labs

October 21, 2010

Laser Nematode Surgery!

Email This Entry

Posted by Derek

There's a headline I've never written before, for sure. A new paper in PNAS describes an assay in nematodes to look for compounds that have an effect on nerve regeneration. That means that you have to damage neurons first, naturally, and doing that on something as small (and as active) as a nematode is not trivial.

The authors (a team from MIT) used microfluidic chips to direct single nematodes into a small chamber where they're held down briefly by a membrane. Then an operator picks out one of its neurons on an imagining screen, whereupon a laser beam cuts it. The nematode is then released into a culture well, where it's exposed to some small molecule to see what effect that has on the neuron's regrowth. It takes about 20 seconds to process a single C. elegans, in case you're wondering, and I can imagine that after a while you'd wish that they weren't streaming along quite so relentlessly.

The group tried about 100 bioactive molecules, targeting a range of known pathways, to see what might speed up or slow down nerve regeneration. As it happens, the highest hit rates were among the kinase inhibitors and compounds targeting cytoskeletal processes. (By contrast, nothing affecting vesicle trafficking or histone deacetylase activity showed any effect). The most significant hit was an old friend to kinase researchers, staurosporine. Interestingly, this effect was only seen on particular subtypes of neurons, suggesting that they weren't picking up some sort of broad-spectrum regeneration pathway.

The paper acknowledges that staurosporine has a number of different activities, but treats it largely as a PKC inhibitor. I'm not sure that that's a good idea, personally - I'd be suspicious of pinning any specific activity to that compound without an awful lot of follow-up, because it's a real Claymore mine when it comes to kinases. The MIT group did check to see if caspases (and apoptotic pathways in general) were involved, since those are well-known effects of staurosporine treatment, and they seem to have ruled those out. And they also followed up with some other PKC inhibitors, chelerythrine and Gö 6983, and these showed similar effects.

So they may be right about this being a PKC pathway, but that's a tough one to nail down. (And even if you do, there are plenty of PKC isoforms doing different things, but there aren't enough selective ligands known to unravel all those yet). Chelerythrine inhibits alanine aminotransferase, has had some doubts expressed about it before in PKC work, and also binds to DNA, which may be responsible for some of its activity in cells. Gö 6983 seems to be a better tool, but it's is in the same broad chemical class as staurosporine itself, so as a medicinal chemist I still find myself giving it the fishy eye.

This is very interesting work, nonetheless, and it's the sort of thing that no one's been able to do before. I'm a big fan of using the most complex systems you can to assay compounds, and living nematodes are a good spot to be in. I'd be quite interested in a broader screen of small molecules, but 20 seconds per nematode surgery is still too slow for the sort of thing a medicinal chemist like me would like to run - a diversity set of, say, ten or twenty thousand compounds, for starters. And there's always the problem we were talking about here the other day, about how easy it is to get compounds into nematodes at all. I wonder if there were some false negatives in this screen just because the critters had no exposure?

Comments (16) + TrackBacks (0) | Category: Biological News | Drug Assays | The Central Nervous System

October 13, 2010

A Cautionary Tale

Email This Entry

Posted by Derek

For those who haven't seen it, I heard Adam Renslo of UCSF present this work yesterday. His group was looking for inhibitors of cruzain, a target for Chagas' disease, which is certainly a worthy cause (and a tough target). They found a series of oxadiazoles, which are, to be sure, rather ugly (but no uglier than a lot of chemical matter you see in the kinase field, among others). They had affinity, they had reasonable SAR, and the team drove the potencies down against the target. . .only to find, late in the game, that it was all an illusion.

These compounds are aggregators. (That link takes you to a post about a follow-up paper from the UCSF folks, covering this debacle and others). What's striking is that this artifact (compound aggregation under the assay conditions) mimicked a plausible SAR - it wasn't just some random thing that made the numbers hard to interpret. No, it looks like Renslo and his team ended up optimizing for aggregation. As he put it in his presentation, "You'll find what you're looking for".

His other quote at the end of the talk was "Small molecules are much stranger than we've been led to believe", and I can't argue with that one either. Before anyone makes a comment about how his group should have checked their assay more thoroughly, or how they shouldn't have been trying to push such an unpleasant-looking series of compounds anyway - in general, about how this wouldn't have happened to you - pause for a moment, and be honest. Renslo was in this paper, and I thank him for it.

Comments (28) + TrackBacks (0) | Category: Drug Assays

October 5, 2010

Chemical Biology: Plastic Antibodies?

Email This Entry

Posted by Derek

Here's an interesting example of a way that synthetic chemistry is creeping into the provinces of molecular biology. There have been a lot of interesting ideas over the years around the idea of polymers made to recognize other molecules. These appear in the literature as "molecularly imprinted polymers", among other names, and have found some uses, although it's still something of a black art. A group at Cal-Irvine has produced something that might move the field forward significantly, though.

In 2008, they reported that they'd made polymer particles that recognized the bee-sting protein melittin. Several combinations of monomers were looked at, and the best seemed to be a crosslinked copolymer with both acrylic acid and an N-alkylacrylamide (giving you both polar and hydrophobic possibilities). But despite some good binding behavior, there are limits to what these polymers can do. They seem to be selective for melittin, but they can't pull it out of straight water, which is a pretty stringent test. (If you can compete with the hydrogen-bonding network of bulk water that's holding the hydrophilic parts of your target, as opposed to relying on just the hydrophobic interactions with the other parts, you've got something impressive).

Another problem, which is shared by all polymer-recognition ideas, is that the materials you produce aren't very well defined. You're polymerizing a load of monomers in the presence of your target molecule, and they can (and will) link up in all sorts of ways. So there are plenty of different binding sites on the particles that get produced, with all sorts of affinities. How do you sort things out?

Now the Irvine group has extended their idea, and found some clever ways around these problems. The first is to use good old affinity chromatography to clean up the mixed pile of polymer nanoparticles that you get at first. Immobilizing melittin onto agarose beads and running the nanoparticles over them washes out the ones with lousy affinity - they don't hold up on the column. (Still, they had to do this under fairly high-salt conditions, since trying this in plain water didn't allow much of anything to stick at all). Washing the column at this point with plain water releases a load of particles that do a noticeably better job of recognizing melittin in buffer solutions.

The key part is coming up, though. The polymer particles they've made show a temperature-dependent change in structure. At RT, they're collapsed polymer bundles, but in the cold, they tend to open up and swell with solvent. As it happens, that process makes them lose their melittin-recognizing abilities. Incubating the bound nanoparticles in ice-cold water seems to only release the ones that were using their specific melittin-binding sites (as opposed to more nonspecific interactions with the agarose and the like). The particles eluted in the cold turned out to be the best of all: they show single-digit nanomolar affinity even in water! They're only a few per cent of the total, but they're the elite.

Now several questions arise: how general is this technique? That is, is melittin an outlier as a peptide, with structural features that make it easy to recognize? If it's general, then how small can a recognition target be? After all, enzymes and receptors can do well with ridiculously small molecules: can we approach that? It could be that it can't be done with such a simple polymer system - but if more complex ones can also be run through such temperature-transition purification cycles, then all sorts of things might be realized. More questions: What if you do the initial polymerization in weird solvents or mixtures? Can you make receptor-blocking "caps" out of these things if you use overexpressed membranes as the templates? If you can get the particles to the right size, what would happen to them in vivo? There are a lot of possibilities. . .

Comments (15) + TrackBacks (0) | Category: Analytical Chemistry | Chemical Biology | Chemical News | Drug Assays

August 25, 2010

GSK's Response to the Sirtuin Critics

Email This Entry

Posted by Derek

OK, time (finally) for the latest chapter in the GSK-Sirtris saga. (This is going to get fairly geeky, so feel free to skip ahead if you're not into enzymology). You'll recall from previous installments that Amgen and Pfizer, among others, had disputed whether the reported sirtuin compounds worked the way that had originally been reported. GSK has now published a paper in the Journal of Biological Chemistry to address those questions. How well does this clear things up? Let's take things in order:

Claim 1: Resveratrol is not a direct activator of SIRT1 activity (Amgen). Building on two 2005 papers, the Amgen team said that resveratrol, the prototype SIRT1 ligand, only works in that manner when the fluorescent peptide (Fluor de Lys) was used in the assay. This is due, they found, exclusively to the fluorophore on the peptide - it's an artifact of the assay conditions. Without it, no activation was seen with protein assays in vitro, nor in cell assays. Native substrates (p53-derived peptide and PGC-1alpha) show nothing.

GSK's response: This is true. They too, found that activation of SIRT1 depends on the structure of the substrate. Without the fluorescent label, no activation is seen.

Claim 2: Not only is this true for resveratrol, it's true for SRT1720, SRT2183, and SRT 1460 (Pfizer). The Pfizer team did a similar breakdown of the assay conditions, and found (through several biophysical methods) that the fluorophore is indeed the crucial element in the activity seen in these assays. And again, since that's an artificial tag, the Fluor de Lys-based assays can have nothing to do with real in vivo activity. Native substrates (p53-derived peptide, full-length p53, and acetyl CoA synthase 1) show nothing.

GSK's response: As above, activation of SIRT1 depends on the structure of the substrate. Without the fluorescent label, no activation is seen. SRT1460 and SRT1720 do indeed bind to the fluorescent peptide, but not to the unlabeled versions. Looking over a broader range of structures, some of them interact with the fluorophore, and some don't. There's no correlation between this affinity and a compound's ability to activate SIRT1.

A screen of 5,000 compounds in this class turned up three that actually do work with nonfluorescent peptide substrates (compounds 22, 23, and 24 in the paper). None of these have been previously disclosed. They, however, that even these still don't work when the peptide substrate lacks both the fluorescent tag and a biotin tag.

What's more, when these three compounds are tested on a p53-derived 20-mer peptide substrate, they actually inhibit acetylation, instead of enhancing it. Looking closer at a range of peptide substrates, SRT1460 and other compounds can also inhibit or enhance acetylation, depending on what peptide is being used. An allosteric mechanism could explain these results. It seems more likely that there are at least two specific sites on SIRT1 that can bind these compounds - the active site and an allosteric one. Thus there are several species in equilibrium, depending on whether these sites have substrate or small molecule bound to them, and on how this binding stabilizes or destabilizes particular pathways. In the real cell, this may all be part of various protein-protein interactions.

Claim 3: SRT1720 does not lower glucose in a high-fat-fed mouse model (Pfizer). Even though exposure of the drug was as reported previously, they saw no evidence (at 30 mg/kilo) of glucose lowering or of any increased mitochondrial function. These animals showed increased food intake and weight gain. The 100 mpk dose was not well tolerated, and killed some animals.

GSK's response: not addressed in this paper. It's an enzymology study only.

Claim 4: Resveratrol, SRT 1460, SRT1720, and SRT2183 are not selective (Pfizer). A screen of over 100 targets showed all of these compounds hitting multiple targets, with resvertrol itself showing the closest thing to a clean profile. None of them, say the Pfizer team, are suitable pharmacological tools.

GSK's response: not addressed in this paper. None of the newly disclosed compounds have selectivity data of this sort attached to them, either. I'd be very curious to know how they look, and I'd be very leery of attaching much importance to their behavior in living systems until that's been done.

The take-home: On the enzymology level, this new paper seems to be solid work. But it's the sort of solid work that should have been done around the time that GSK bought Sirtris, and not something appearing in 2010 in response to major attacks in the literature. The first main claim of those attacking papers is, in fact, absolutely true: the original Fluor de Lys assay is worthless for characterizing these compounds. What we learn from this paper is that the assay is worthless for even more complicated reasons than originally thought, and that the whole series of SRT compounds behaves in ways that were not apparent from the published work, to put it lightly.

As to the selectivity and in vivo effects of these compounds, Pfizer's gauntlet is still thrown down right where they left it. The fact that these compounds are so much harder to understand than was originally thought, even in well-controlled enzyme assays, makes me wonder how easy it will be to figure out the rest of the story. . .

Comments (37) + TrackBacks (0) | Category: Aging and Lifespan | Drug Assays | Drug Development

July 12, 2010

Natural Products: Not the Best Fit for Drugs?

Email This Entry

Posted by Derek

Stuart Schreiber and Paul Clemons of the Broad Institute have a provocative paper out in JACS on natural products and their use in drug discovery. As many know, a good part of the current pharmacopeia is derived from natural product lead structures, and in many other cases a natural product was essential for identifying a target or pathway for a completely synthetic compound.

But are there as many of these cases as we think - or as there should be? This latest paper takes a large set of interaction data and tries to map natural product activities on to it. It's already know that there are genes all up and down the "interactome" spectrum, as you'd expect, with some that seem to be at the crossroads of dozens (or hundreds) of pathways, and others that are way out on the edges. And it's been found that disease targets tend to fall in the middle of this range, and not so much in the too-isolated or too-essential zones on either side.

That seems reasonable. But then comes the natural product activity overlay, and there the arguing can start. Natural products, the paper claims, tend to target the high-interaction essential targets at the expense of more specific disease targets. They're under-represented in the few-interaction group, and very much over-represented in the higher ones. Actually, that actually seems reasonable, too - most natural products are produced by organisms as essentially chemical warfare, and the harder they can hit, the better. Looking at subsets of the natural product list (only the most potent compounds, for example) did not make this effect vanish. Meanwhile, if you look at the list of approved drugs (minus the natural products on it), that group fits the middle-range interactivity group much more closely.

But what does that mean for natural products as drug leads? There would appear to be a mismatch here, with a higher likelihood of off-target effects and toxicity among a pure natural-product set. (The mismatch, to be more accurate, is between what we want exogenous chemicals to do versus what evolution has selected them to do). The paper ends up pointing out that additional sources of small molecules look to be needed outside of natural products themselves.

I'll agree with that. But I suspect that I don't agree with the implications. Schreiber has long been a proponent of "diversity-oriented synthesis" (DOS), and would seem to be making a case for it here without ever mentioning it by name. DOS is the idea of making large collections of very structurally diverse molecules, with an eye to covering as much chemical space as possible. My worries (expressed in that link above) are that the space it covers doesn't necessarily overlap very well with the space occupied by potential drugs, and that chemical space is too humungously roomy in any event to be attacked very well by brute force.

Schreiber made a pitch a few years ago for the technique, that time at the expense of small-molecule compound collections. He said that these were too simple to hit many useful targets, and now he's taking care of the natural product end of the spectrum by pointing out that they hit too many. DOS libraries, then, must be just in the right range? I wish he'd included data on some of them in this latest paper; it would be worthwhile to see where they fell in the interaction list.

Comments (58) + TrackBacks (0) | Category: Drug Assays | Drug Industry History | Toxicology

April 27, 2010

Masses of Data, In Every Sample

Email This Entry

Posted by Derek

I've said several times that I think that mass spectrometry is taking over the analytical world, and there's more evidence of that in Angewandte Chemie. A group at Justus Liebig University in Giessen has built what has to be the finest imaging mass spec I've ever seen. It's a MALDI-type machine, which means that a small laser beam does the work of zapping ions off the surface of the sample. But this one has better spatial resolution than anything reported so far, and they've hooked it up to a very nice mass spec system on the back end. The combination looks to me like something that could totally change the way people do histology.

For the non-specialist readers in the audience, mass spec is a tremendous workhorse of analytical chemistry. Basically, you use any of a whole range of techniques (lasers, beams of ions, electric charges, etc.) to blast individual molecules (or their broken parts!) down through a chamber and determine how heavy each one is. Because molecular weights are so precise, this lets you identify a lot of molecules by both their whole weights - their "molecular ions" - and by their various fragments. Imagine some sort of crazy disassembler machine that rips things - household electronic gear, for example - up into pieces and weighs every chunk, occasionally letting a whole untouched unit through. You'd see the readouts and say "Ah-hah! Big one! That was a plasma TV, nothing else is up in that weight range. . .let's see, that mix of parts coming off it means that it must have been a Phillips model so-and-so; they always break up like that, and this one has the heavier speakers on it." But mass spec isn't so wasteful, fortunately: it doesn't take much sample, since there are such gigantic numbers of molecules in anything large enough to see or weigh.
Take a look at this image. That's a section of a mouse pituitary gland - on the right is a standard toluidine-blue stain, and on the left is the same tissue slice as imaged (before staining) by the mass spec. The green and blue colors are two different mass peaks (826.5723 and 848.5566, respectively), which correspond to different types of phospholipid from the cell membranes. (For more on such profiling, see here). The red corresponds to a mass peak for the hormone vasopressin. Note that the difference in phospholipid peaks completely shows the difference between the two lobes of the gland (and also shows an unnamed zone of tissue around the posterior lobe, which you can barely pick up in the stained preparation). The vasopressin is right where it's supposed to be, in the center of the posterior lobe.

One of the most interesting things about this technique is that you don't have to know any biomarkers up front. The mass spec blasts away at each pixel's worth of data in the tissue sample and collects whatever pile of varied molecular-weight fragments that it can collect. Then the operator is free to choose ions that show useful contrasts and patterns (I can imagine software algorithms that would do the job for you - pick two parts of an image and have the machine search for whatever differentiates them). For instance, it's not at all clear (yet) why those two different phospholipid ions do such a good job at differentiating out the pituitary lobes - what particular phospholipids they correspond to, why the different tissues have this different profile, and so on. But they do, clearly, and you can use that to your advantage.

As this technique catches on, I expect to see large databases of mass-based "contrast settings" develop as histologists find particularly useful readouts. (Another nice feature is that one can go back to previously collected data and re-process for whatever interesting things are discovered later on). And each of these suggests a line of research all its own, to understand why the contrast exists in the first place.
The second image shows ductal carcinoma in situ. On the left is an optical image, and about all you can say is that the darker tissue is the carcinoma. The right-hand image is colored by green (mass of 529.3998) and red (mass of 896.6006), which correspond to healthy and cancerous tissue, respectively (and again, we don't know why, yet). But look closely and you can see that some of the dark tissue in the optical image doesn't actually appear to be cancer - and some of dark spots in the lighter tissue are indeed small red cells of trouble. We may be able to use this technology to diagnose cancer subtypes more accurately than ever before - the next step will be to try this on a number of samples from different patients to see how much these markers vary. I also wonder if it's possible to go back to stored tissue samples and try to correlate mass-based markers with the known clinical outcomes and sensitivities to various therapies.

I'd also be interested in knowing if this technique is sensitive enough to find small-molecule drugs after dosing. Could we end up doing pharmacokinetic measurements on a histology-slide scale? Ex vivo, could we possibly see uptake of our compounds once they're applied to a layer of cells in tissue culture? Oh, mass spec imaging has always been a favorite of mine, and seeing this level of resolution just brings on dozens of potential ideas. I've always had a fondness for label-free detection techniques, and for methods that don't require you to know too much about the system before being able to collect useful data. We'll be hearing a lot more about this, for sure.

Update: I should note that drug imaging has certainly been accomplished through mass spec, although it's often been quite the pain in the rear. It's clearly a technology that's coming on, though.

Comments (9) + TrackBacks (0) | Category: Analytical Chemistry | Biological News | Cancer | Drug Assays

April 26, 2010

Charles River Buys WuXi

Email This Entry

Posted by Derek

I don't think we saw this one coming: Charles River Labs has announced that they're buying WuXi PharmaTech. They're paying about a 28% premium over Friday's closing stock price - Charles River's CEO will stay on, and WuXi's founder (Li Ge) will serve as executive VP under him.

Charles River, which is strong in the animal-testing end of the business, has apparently decided that Wu Xi is one of their biggest competitors (I'd agree) and has decided to try to stake out a leading position in the whole contract-research space. It's interesting to me that the folks at Wu Xi bought into this reasoning as well, although (since they're a publicly traded company here in the US), a lucrative stock offer can be its own argument. One now wonders, though, about the company's statements on re-staffing some of their US labs when economic conditions improve. . .

Comments (15) + TrackBacks (0) | Category: Animal Testing | Business and Markets | Drug Assays | Drug Development

April 6, 2010

Take These?

Email This Entry

Posted by Derek

A reader sends along a note about this patent application from the University of Rochester. The inventor, David Goldfarb, seems to have used an assay (the subject of a previous application) to screen a library of commercially available compounds for potential life-extending properties in model organisms. Here's some detail on the screen from PubChem.

The abstract of the application makes it sound worse than it is: "A method for altering the lifespan of a eukaryotic organism. The method comprises the steps of providing a lifespan altering compound, and administering an effective amount of the compound to a eukaryotic organism, such that the lifespan of the organism is altered. . ." That sounds like one of those "Oh, get real" applications that the patent databases are cluttered with. But when you get to the claims, you find that a list of compounds is specifically given, with more- and most-preferred ones as you go down. And I don't have a problem with that, as far as it goes - the inventor has an assay, has run a bunch of compounds through it, and finds that some of them have utility that apparently no one else has recognized.

The compounds themselves, though. . .well, here are the specifically claimed ones on the list. I don't necessarily see aliphatic triketones extending my life, but perhaps I'm cynical.

Comments (18) + TrackBacks (0) | Category: Aging and Lifespan | Drug Assays | Patents and IP

March 30, 2010

Animal Studies: Are Too Many Never Published At All?

Email This Entry

Posted by Derek

A new paper in PLoS Biology looks at animal model studies reported for the treatment of stroke. The authors use statistical techniques to try to estimate how many have gone unreported. From a database with 525 sources, covering 16 different attempted therapies (which together come to 1,359 experiments and 19,956 animals), they find that only a very small fraction of the publications (about 2%) report no significant effects, which strongly suggests that there is a publication bias at work here. The authors estimate that there may well be around 200 experiments that showed no significant effect and were never reported, whose absence would account for around one-third of the efficacy reported across the field. In case you're wondering, the therapy least affected by publication bias was melatonin, and the one most affected seems to be administering estrogens.

I hadn't seen this sort of study before, and the methods they used to arrive at these results are interesting. If you plot the precision of the studies (Y axis) versus the effect size (X axis), you should (in theory) get a triangular cloud of data. As the precision goes down, the spread of measurements across the X-axis increases, and as the precision goes up, the studies should start to converge on the real effect of the treatment, whatever that might be. (In this study, the authors looked only at reported changes in infarct size as a measure of stroke efficacy). But in many of the reported cases, the inverted-funnel shape isn't symmetrical - and every single time that happens, it turns out that the gaps are in the left-hand side of the triangle, the not-as-precise and negative-effect regions of the plots. This doesn't appear to be just due to less-precise studies tending to show positive effects for some reason - it strongly suggests that there are negative studies that just haven't been reported.

The authors point out that applying their statistical techniques to reported human clinical studies is more problematic, since smaller (and thus less precise) trials may well involve unrepresentative groups of patients. But animal studies are much less prone to this problem.

The loss of experiments that showed no effect shouldn't surprise anyone - after all, it's long been known that publishing such papers is just plain harder than publishing ones that show something happening. There's an obvious industry bias toward only showing positive data, but there's an academic one, too, which affects basic research results. As the authors put it:

These quantitative data raise substantial concerns that publication bias may have a wider impact in attempts to synthesise and summarise data from animal studies and more broadly. It seems highly unlikely that the animal stroke literature is uniquely susceptible to the factors that drive publication bias. First, there is likely to be more enthusiasm amongst scientists, journal editors, and the funders of research for positive than for neutral studies. Second, the vast majority of animal studies do not report sample size calculations and are substantially underpowered. Neutral studies therefore seldom have the statistical power confidently to exclude an effect that would be considered of biological significance, so they are less likely to be published than are similarly underpowered “positive” studies. However, in this context, the positive predictive value of apparently significant results is likely to be substantially lower than the 95% suggested by conventional statistical testing. A further consideration relating to the internal validity of studies is that of study quality. It is now clear that certain aspects of experimental design (particularly randomisation, allocation concealment, and the blinded assessment of outcome) can have a substantial impact on the reported outcome of experiments. While the importance of these issues has been recognised for some years, they are rarely reported in contemporary reports of animal experiments.

And there's an animal-testing component to these results, too, of course. But lest activists seize on the part of this paper that suggests that some animal testing results are being wasted, they should consider the consequences (emphasis below mine):

The ethical principles that guide animal studies hold that the number of animals used should be the minimum required to demonstrate the outcome of interest with sufficient precision. For some experiments, this number may be larger than those currently employed. For all experiments involving animals, nonpublication of data means those animals cannot contribute to accumulating knowledge and that research syntheses are likely to overstate biological effects, which may in turn lead to further unnecessary animal experiments testing poorly founded hypotheses.

This paper is absolutely right about the obligation to have animal studies mean something to the rest of the scientific community, and it's clear that this can't happen if the results are just sitting on someone's hard drive. But it's also quite possible that for even some of the reported studies to have meant anything, that they would have had to have used more animals in the first place. Nothing's for free.

Comments (19) + TrackBacks (0) | Category: Animal Testing | Cardiovascular Disease | Clinical Trials | Drug Assays | The Scientific Literature

March 29, 2010

Compounds and Proteins

Email This Entry

Posted by Derek

For the medicinal chemists in the audience, I wanted to strongly recommend a new paper from a group at Roche. It's a tour through the various sorts of interactions between proteins and ligands, with copious examples, and it's a very sensible look at the subject. It covers a number of topics that have been discussed here (and throughout the literature in recent years), and looks to be an excellent one-stop reference.

In fact, read the right way, it's a testament to how tricky medicinal chemistry is. Some of the topics are hydrogen bonds (and why they can be excellent keys to binding or, alternatively, of no use whatsoever), water molecules bound to proteins (and why disturbing them can account for large amounts of binding energy, or, alternatively, kill your compound's chances of ever binding at all), halogen bonds (which really do exist, although not everyone realizes that), interactions with aryl rings (some of which can be just as beneficial coming in 90 degrees to where you might imagine), and so on.

And this is just to get compounds to bind to their targets, which is the absolute first step on the road to a drug. Then you can start worrying about how to have your compounds not bind to things you don't want (many of which you probably don't even realize even are out there). And about how to get it to decent blood levels, for a decent amount of time, and into the right compartments of the body. And at that point, it's nearly time to see if it does any good for the disease you're trying to target!

Comments (5) + TrackBacks (0) | Category: Drug Assays | In Silico | Life in the Drug Labs

March 26, 2010

Privileged Scaffolds? How About Unprivileged Ones?

Email This Entry

Posted by Derek

The discussion of "privileged scaffolds" in drugs here the other day got me to thinking. A colleague of mine mentioned that there may well be structures that don't hit nearly as often as you'd think. The example that came to his mind was homopiperazine, and he might have a point; I've never had much luck with those myself. That's not much of a data set, though, so I wanted to throw the question out for discussion.

We'll have to be careful to account for Commercial Availability Bias (which at least for homopiperazines has decreased over the years) and Synthetic Tractability Bias. Some structures don't show up much because they just don't get made much. And we'll also have to be sure that we're talking about the same things: benzo-fused homopiperazines (and other fused seven-membered rings) hit like crazy, as opposed to the monocyclic ones, which seem to be lower down the scale, somehow.

It's not implausible that there should be underprivileged scaffolds. The variety of binding sites is large, but not infinite, and I'm sure that it follows a power-law distribution like so many other things. The usual tricks (donor-acceptor pairs spaced about so wide apart, pi-stacking sandwiches, salt bridges) surely account for much more than their random share of the total amount of binding stabilization out there in the biosphere. And some structures are going to match up with those motifs better than others.

So, any nominations? Have any of you had structural types that seem as if they should be good, but always underperform?

Comments (9) + TrackBacks (0) | Category: Drug Assays | Drug Industry History | Life in the Drug Labs

March 24, 2010

Privileged Scaffolds

Email This Entry

Posted by Derek

Here's a new article on the concept of "privileged scaffolds", the longstanding idea that there seem to be more biologically active compounds built around some structures than others. This doesn't look like it tells me anything I didn't know, but it's a useful compendium of such structures if you're looking for one. Overall, though, I'm unsure of how far to push this idea.

On the one hand, it's certainly true that some structural motifs seem to match up with binding sites more than others (often, I'd say, because of some sort of donor-acceptor pair motif that tends to find a home inside protein binding sites). But in other cases, I think that the appearance of what looks like a hot scaffold is just an artifact of everyone ripping off something that worked - others might have served just as well, but people ran with what had been shown to work. And then there are other cases, where I think that the so-called privileged structure should be avoided for everyone's good: our old friend rhodanine makes an appearance in this latest paper, for example. Recall this this one has been referred to as "polluting the literature", with which judgment I agree.

Comments (11) + TrackBacks (0) | Category: Drug Assays | Drug Industry History

Drugs And Their Starting Points

Email This Entry

Posted by Derek

I've spoken about fragment-based drug design and ligand efficiency here a few times. There's a new paper in J. Med. Chem. that puts some numbers on that latter concept. (Full disclosure - I've worked with its author, although I had nothing to do with this particular paper).

For the non-chemists in the crowd who want to know what I'm talking about, fragment-based methods are an attempt to start with smaller, weaker-binding chemical structures than we usually work with. But if you look at how much affinity you're getting for the size of the molecules, you find that some of these seemingly weaker compounds are actually doing a great job for their size. Starting from these and building out, with an eye along the way toward keeping that efficiency up, could be a way of making better final compounds than you'd get by starting from something larger.

Looking over a number of examples where the starting compounds can be compared to the final drugs (not a trivial data set to assemble, by the way), this work finds that drugs, compared to their corresponding leads, tend to have similar to slightly higher binding efficiencies, although there's a lot of variability. They also tend to have similar logP values, which is a finding that doesn't square with some previous analyses (which showed things getting worse during development). But drugs are almost invariably larger than their starting points, so no matter what, one of the keys is not to make the compounds greasier as you add molecular weight. (My "no naphthyls" rule comes from this, actually).

There are a few examples of notably poor ligand-efficient starting structures that have nonetheless been developed into drugs. Interestingly, several of these are the HIV protease inhibitors, with Reyataz (atazanavir) coming in as the least ligand-efficient drug in the whole data set. A look at its structure will suffice. The wildest one on the list appears to be no-longer-marketed amprenavir, whose original lead was 53 micromolar and weighed over 600, nasty numbers indeed. I would not recommend emulating that one. In case you're wondering, the most ligand efficient drug in the set is Chantix (varenicline).

In the cases where ligand efficiency actually went down along the optimization route, inspection of the final structures shows that in many cases, the discovery team was trading efficiency for some other property (PK, solubility, etc.) To me, that's another good argument to make things as efficient as you can, because that gives you something to trade. A big, chunky, lashed-together structure doesn't give you much room to maneuver.

Comments (27) + TrackBacks (0) | Category: Drug Assays | Drug Development

March 23, 2010

We Don't Know Beans About Biotin

Email This Entry

Posted by Derek

You know, you'd think that we'd understand the way things bind to proteins well enough to be able to explain why biotin sticks so very, very tightly to avidins. That's one of the most impressive binding events in all of biology, short of pushing electrons and forming a solid chemical bond - biotin's stuck in there at femtomolar levels. It's so strong and so reliable that this interaction is the basis for untold numbers of laboratory and commercial assays - just hang a biotin off one thing, expose it to something else that has an avidin (most often streptavidin) coated on it, and it'll stick, or else something is Very Wrong. So we have that all figured out.

Wrong. Turns out that there's a substantial literature given to arguing about just why this binding is so tight. One group holds out for hydrophobic interactions (which seems rather weird to me, considering that biotin's rather polar by most standards). Another group has a hydrogen-bonding explanation, which (on the surface) seems more feasible to me. Now a new paper says that the computational methods applied so far can't handle electrostatic factors well, and that those are the real story.

I'm not going to take a strong position on any of these; I'll keep my head down while the computational catapults launch at each other. But it's definitely worth noting that we apparently can't explain the strongest binding site interaction that we know of. It's the sort of thing that we'd all like to be able to generate at will in our med-chem programs, but how can we do that when we don't even know what's causing it?

Comments (12) + TrackBacks (0) | Category: Drug Assays | In Silico

March 15, 2010

Tricor's Troubles

Email This Entry

Posted by Derek

It's easy to lose sight of what a drug is supposed to do. Many conditions come on so slowly that we have to use blood chemistry or other markers to see the progress of therapy in a realistic time. And over time, that blood marker can get confused with the disease itself.

To pick one famous example, try cholesterol. Everyone you stop on the street will know that "high cholesterol is bad for you". But the first thing you have to do is distinguish between LDL and HDL cholesterol - if the latter is a large enough fraction of the total, the aggregate number doesn't matter as much. And fundamentally, there's not a disease called "high cholesterol" - that's a symptom of some other cluster of metabolic processes that have gone subtly off. And the endpoint of any therapy in that field isn't really to lower the number in a blood test: it's to prevent heart attacks and to extend healthy lifetimes, mortality and morbidity. As we're seeing with Vytorin, it may be possible to drop the numbers in a blood test but not see the benefit that's supposed to be there.

Another example of this came up over the weekend. The fibrates are a class of drugs that change lipid levels, although the way they work is still rather obscure. They're supposed to be ligands for the PPAR-alpha nuclear receptor, but they're not very potent against it when you study that closely. At any rate, they do lower triglycerides and have some other effects, which should be beneficial in patients whose lipids are off and are at risk for cardiac problems.

But are they? Type II diabetics tend to be people who fit that last category well, and that's where a lot of fenofibrate is prescribed (as Abbott's Tricor in the US, and under a number of other names around the world). A five-year study in over five thousand diabetic patients, though, has just shown no difference versus placebo. Again, there's no doubt that the drug lowers triglycerides and changes the HDL/LDL/VLDL ratios. It's just that, for reasons unknown, doing so with fenofibrate doesn't seem to actually help diabetic patients avoid cardiac trouble.

Mortality and morbidity: lowering them is a very tough test for any drug, but if you can't, then what's the point of taking something in the first place? This is something to keep in mind as the push for biomarkers delivers more surrogate endpoints. Some of them will, inevitably, turn out not to mean as much as they're supposed to mean.

Comments (15) + TrackBacks (0) | Category: Cardiovascular Disease | Clinical Trials | Diabetes and Obesity | Drug Assays

March 12, 2010

Garage Biotech

Email This Entry

Posted by Derek

Freeman Dyson has written about his belief that molecular biology is becoming a field where even basement tinkerers can accomplish things. Whether we're ready for it or not, biohacking is on its way. The number of tools available (and the amount of surplus equipment that can be bought) have him imagining a "garage biotech" future, with all the potential, for good and for harm, that that entails.

Well, have a look at this garage, which is said to be somewhere in Silicon Valley. I don't have any reason to believe the photos are faked; you could certainly put your hands on this kind of equipment very easily in the Bay area. The rocky state of the biotech industry just makes things that much more available. From what I can see, that's a reasonably well-equipped lab. If they're doing cell culture, there needs to be some sort of incubator around, and presumably a -80 degree freezer, but we don't see the whole garage, do we? I have some questions about how they do their air handling and climate control (although that part's a bit easier in a California garage than it would be in a Boston one). There's also the issue of labware and disposables. An operation like this does tend to run through a goodly amount of plates, bottles, pipet tips and so on, but I suppose those are piled up on the surplus market as well.

But what are these folks doing? The blog author who visited the site says that they're "screening for anti-cancer compounds". And yes, it looks as if they could be doing that, but the limiting reagent here would be the compounds. Cells reproduce themselves - especially tumor lines - but finding compounds to screen, that must be hard when you're working where the Honda used to be parked. And the next question is, why? As anyone who's worked in oncology research knows, activity in a cultured cell line really doesn't mean all that much. It's a necessary first step, but only that. (And how many different cell lines could these people be running?)

The next question is, what do they do with an active compound when they find one? The next logical move is activity in an animal model, usually a xenograft. That's another necessary-but-nowhere-near-sufficient step, but I'm pretty sure that these folks don't have an animal facility in the basement, certainly not one capable of handling immunocompromised rodents. So put me down as impressed, but puzzled. The cancer-screening story doesn't make sense to me, but is it then a cover for something else? What?

If this post finds its way to the people involved, and they feel like expanding on what they're trying to accomplish, I'll do a follow-up. Until then, it's a mystery, and probably not the only one of its kind out there. For now, I'll let Dyson ask the questions that need to be asked, from that NYRB article linked above:

If domestication of biotechnology is the wave of the future, five important questions need to be answered. First, can it be stopped? Second, ought it to be stopped? Third, if stopping it is either impossible or undesirable, what are the appropriate limits that our society must impose on it? Fourth, how should the limits be decided? Fifth, how should the limits be enforced, nationally and internationally? I do not attempt to answer these questions here. I leave it to our children and grandchildren to supply the answers.

Comments (42) + TrackBacks (0) | Category: Biological News | Drug Assays | General Scientific News | Regulatory Affairs | Who Discovers and Why

March 2, 2010

Why You Don't Want to Make Death-Star-Sized Drugs

Email This Entry

Posted by Derek

I was just talking about greasy compounds the other day, and reasons to avoid them. Right on cue, there's a review article in Expert Opinion in Drug Discovery on lipophilicity. It has some nice data in it, and I wanted to share a bit of it here. It's worth noting that you can make your compounds too polar, as well as too greasy. Check these out - the med-chem readers will find them interesting, and who knows, others might, too:
So, what are these graphs? They show how well compound cross the membranes of Caco-2 cells, a standard assay for permeability. These cells (derived from human colon tissue) have various active-transport pumps going (in both directions), and you can grow them in a monolayer, expose one side to a solution of drug substance, and see how much compound appears on the other side and how quickly. (Of course, good old passive diffusion is also operating, too - a lot of compounds cross membranes by just soaked on through them).

Now, I have problems with extrapolating Caco-2 data too vigorously to the real world - if you have five drug candidates from the same series and want to rank order them, I'd suggest getting real animal data rather than rely on the cell assay. The array of active transport systems (and their intrinsic activity) may well not match up closely enough to help you - as usual, cultured cell lines don't necessarily match reality. But as a broad measure of whether a large set of compounds has a reasonable chance of getting through cell membranes, the assay's not so bad.

First, we have a bunch of compounds with molecular weights between 350 and 400 (a very desirable space to occupy). The Y axis is the partitioning between the two sides of the cells, and X axis is LogD, a standard measure of compound greasiness. That thin blue line is the cutoff for 100 nanomoles/sec of compound transport, so the green compounds above it travel across the membrane well, and the red ones below it don't cross so readily. You'll note that as you go to the left (more and more polar, as measured by logD), the proportion of green compounds gets smaller and smaller. They're rather hang out in the water than dive through any cell membranes, thanks.

So if you want a 50% chance of hitting that 100 nm/sec transport level, then you don't want to go much more polar than a LogD of 2. But that's for compounds in the 350-400 weight range - how about the big heavyweights? Those are shown in the second graph, for compounds greater than 500. Note that the distribution has scrunched disturbingly. Now almost everything is lousy, and if you want that 50% chance of good penetration, you're going to have to get up to a logD of at least 4.5.

That's not too good, because you're always fighting a two-front war here. If you make your compounds that greasy (or more) to try to improve their membrane-crossing behavior, you're opening yourself up (as I said the other day) to more metabolic clearance and more nonspecific tox, as your sticky compounds glop onto all sorts of things in vivo. (They'll be fun to formulate, too). Meanwhile, if you dip down too far into that really-polar left-hand side, crossing your fingers for membrane crossing, you can slide into the land of renal clearance, as the kidneys vacuum out your water-soluble wonder drug and give your customers very expensive urine.

But in general, you have more room to maneuver in the lower molecular weight range. The humungous compounds tend to not get through membranes at reasonable LogD values. And if you try to fix that by moving to higher LogD, they tend to get chewed up or do unexpectedly nasty things in tox. Stay low and stay happy.

Comments (24) + TrackBacks (0) | Category: Drug Assays | Pharma 101 | Pharmacokinetics

March 1, 2010

Calorimetry: What Say You?

Email This Entry

Posted by Derek

I've been involved in a mailing list discussion that I wanted to open up to a wider audience in drug discovery, so here goes. We spend our time (well, a lot of it, when we're not filling out forms) trying to get compound to bind well to our targets. And that binding is, of course, all about energy: the lower the overall energy of the system when your compound binds, relative to the starting state, the tighter the binding.

That energy change can be broken down (all can all chemical free energy changes) into an enthalpic part and an entropic part (that latter one depends on temperature, but we'll assume that everything's being done at a constant T and ignore that part). Roughly speaking, the enthalpic component is where you see effects of hydrogen bonds, pi-pi stacking, and other such "productive" interactions, and the entropic part is where you're pushing water molecules and side chains around - hydrophobic interactions and such.

That's a gross oversimplification, but it's a place to start. It's important to remember that these things are all tangled together in most cases. If you come in with a drug molecule and displace a water molecule that was well-attached to your binding pocket, you've broken some hydrogen bonds - for which you'll pay in enthalpy. But you may well have formed some, too, to your molecule - so you'll get some enthalpy term back. And by taking a bound water and setting it free, you'll pick up some good entropy change, too. But not all waters are so tightly bound - there are a few cases where they're actually at a lower entropy state in a protein pocket then they are out in solution, so displacing one of those actually hurts you in entropy. Hmm.

And as I mentioned here, you have the motion of your drug molecule to consider. If it goes from freely rotating to stuck when it binds (as it may well), then you're paying entropy costs. (That's one reason why tying down your structure into a ring can help so dramatically, when it helps at all). And don't forget the motion of the protein overall - if it's been flopping around until it folds over and clenches down on your molecule, there's another entropy penalty for you, which you'd better be able to make up in enthalpy. And so on.

There's been a proposal, spread most vigorously by Ernesto Freire of Johns Hopkins, that drug researchers should use calorimetry to pick compounds that have the biggest fraction of their binding from enthalpic interactions. (That used to be a terrible pain to do, but recent instruments have made it much more feasible). His contention is that the "best in class" drugs in long-lived therapeutic categories tend to move in that direction, and that we can use this earlier in our decision-making process. People doing fragment-based drug discovery are also urged to start with enthalpically-biased fragments, so that the drug candidate that grows out from them will have a better chance of ending up in the same category.

One possible reason for all this is that drugs that get most of their binding from sheer greasiness, fleeing the water to dive into a protein's sheltering cave, might not be so picky about which cave they pick. There's a persistent belief, which I think is correct, that very hydrophobic compounds tend to have tox problems, because they're often just not selective enough about where they bind. And then they tend to get metabolized and chewed up more, too, which adds to the problem.

And all that's fine. . .except for one thing: is anyone actually doing this? That's the question that came up recently, and (so far), for what it's worth, no one's willing to speak up and say that they are. Perhaps all this is a new enough consideration that all the work is still under wraps. But it will be interesting to see if it holds up or not. We need all the help we can get in drug discovery, so if this is real, then it's welcome. But we also don't need to run more assays that only confuse things, either, so it would be worth knowing if drug-candidate calorimetry falls into that roomy category, too. Opinions?

Comments (26) + TrackBacks (0) | Category: Analytical Chemistry | Drug Assays

February 19, 2010

Two For One Sale

Email This Entry

Posted by Derek

A double complaint this morning, and both from the same literature item - if I were charging anything for the blog, I'd say that it's delivering value for the money. At any rate, the first kvetch is something that I know that many chemists have noticed when reading more biology/medical-oriented journals. You'll see some paper that talks about a new compound that does X, Y, and Z. It'll be named with some sort of code, and they'll tell you all about its interesting effects. . .but they don't get around to actually telling you what the damned stuff is.

As I say, this is a chemist's complaint. Many biologists are fine stipulating that there's a compound that will do these interesting things, because they're mostly interested in hearing about the interesting things themselves. It could just be Compound X as far as they're concerned. But chemists want to see what kind of structure it is that causes all these publication-worthy results, and sometimes we go away disappointed.

Or we have to dig. Take this PNAS paper on a broad-spectrum antiviral compound, LJ001. It looks quite interesting, with effects on a number of different viral types, and through a unique mechanism that targets viral membranes. But what is it? You'll look in vain through the whole paper to find out - that compound is LJ001 to you, Jack. You have to go to the supplemental material to find out, and to page 10 at that.

And that brings up the second complaint. LJ001 turns out to be a rhodanine, and regular readers will note that earlier this month some time was spent here talking about just how ugly and undesirable those are. It's very, very hard to get anyone in the drug business to take a compound in that class seriously, because they have such a poor track record. Looking over the small SAR table provided, I note that if you switch that thioamide group (the part that the chemists hate the most) to a regular amide, turning the thing into an thiazolidinedione, you lose all the activity.

TZDs aren't everyone's favorite group, but at least they've made it into marketed drugs. Rhodanines are no one's favorite group, and it would be a good thing of the authors of these papers would realize that, or at least acknowledge it if they do. It's not an irrational prejudice.

Comments (32) + TrackBacks (0) | Category: Drug Assays | The Scientific Literature

February 8, 2010

Polluting the Literature with PAINs

Email This Entry

Posted by Derek

There's an article out from a group in Australia on the long-standing problem of "frequent hitter" compounds. Everyone who's had to work with high-throughput screening data has had to think about this issue, because it's clear that some compounds are nothing but trouble. They show up again and again as hits in all sorts of assays, and eventually someone gets frustrated enough to flag them or physically remove them from the screening deck (although that last option is often a lot harder than you'd think, and compound flags can proliferate to the point that they get ignored).

The larger problem is whether there are whole classes of compounds that should be avoided. It's not an easy one to deal with, because the question turns on how you're running your assays. Some things are going to interfere with fluorescent readouts, by absorbing or emitting light of their own, but that can depend on the wavelengths you're using. Others will mung up a particular coupled assay readout, but leave a different technology untouched.

And then there's the aggregation problem, which we've only really become aware of in the past few years. Some compounds just like to stick together into huge clumps, often taking the assay's protein target (or some other key component) with them. At first, everyone thought "Ah-hah! Now we can really scrub the screening plates of all the nasties!", but it turns out that aggregation itself is an assay-dependent phenomenon. Change the concentrations or added proteins, and whoomph: compounds that were horrible before suddenly behave reasonably, while a new set of well-behaved structures has suddenly gone over to the dark side.

This new paper is another attempt to find "Pan-Assay Interference" compounds or PAINs, as they name them. (This follows a weird-acronym tradition in screening that goes back at least to Vertex's program to get undesirable structures out of screening collections, REOS, for "Rapid Elimination of, uh, Swill"). It will definitely be of interest to people using the AlphaScreen technology, since it's the result of some 40 HTS campaigns using it, but the lessons are worth reading about in general.

What they found was that (as you'd figure) that while it's really hard to blackball compounds permanently with any degree of confidence, the effort needs to be made. Still, even using their best set of filters, 5% of marketed drugs get flagged as problematic screening hits - in fact, hardly any database gives you a warning rate below that, with the exception of a collection of CNS drugs, whose properties are naturally a bit more constrained. Interestingly, they also report the problematic-structure rate for the collections of nine commercial compound vendors, although (frustratingly) without giving their names. Several of them sit around that 5% figure, but a couple of them stand out with 11 or 12% of their compounds setting off alarms. This, the authors surmise, is linked to some of the facile combinatorial-type reactions used to prepare them, particularly ones that leave enones or exo-alkenes in the final structures.

So what kinds of compounds are the most worrisome? If you're going to winnow out anything, you should probably start with these: Rhodanines are bad, which doesn't surprise me. (Abbott and Bristol Myers-Squibb have also reported them as troublesome). Phenol Mannich compounds and phenolic hydrazones are poor bets. And all sort of keto-heterocycles with conjugated exo alkenes make the list. There are several other classes, but those are the worst of the bunch, and I have to say, I'd gladly cross any of them off a list of screening hits.

But not everyone does. As the authors show, there are nearly 800 literature references to rhodanine compounds showing biological effects. A conspicuous example is here, from the good folks at Harvard, which was shown to be rather nonspecifically ugly here. What does all this do for you? Not much:

"Rather than being privileged structures, we suggest that rhodanines are polluting the scientific literature. . .these results reflect the extent of wasted resources that these nuisance compounds are generally causing. We suggest that a significant proportion of screening-based publications and patents may contain assay interference hits and that extensive docking computations and graphics that are frequently produced may often be meaningless. In the case of rhodanines, the answer set represents some 60 patents and we have found patents to be conspicuously prevalent for other classes of PAINS. This collectively represents an enormous cost in protecting intellectual property, much of which may be of little value. . ."

Comments (11) + TrackBacks (0) | Category: Drug Assays | Drug Industry History | The Scientific Literature

January 29, 2010

Merck and Sirna

Email This Entry

Posted by Derek

Xconomy has a look inside the Merck-Sirna acquisition, an interview with Merck's head of that area. As you'd guess, he emphasizes that one of the biggest challenges in the field is delivery, and he makes the pitch that this is how Merck is going to make this work out:

What you often read about, but many people don’t understand, is how hard it is to make a drug. Our approach to RNA Therapeutics is made with a recognition of the full package it takes to launch a successful commercial product. . .That’s versus another strategy you see from smaller companies, which is to get an interesting experimental result, and publicly disclose it in an attempt to increase the value of your investment or a VC’s investment, without a real [awareness] of what it will take to make a therapeutic eight years later. . .

We immediately, after the acquisition, invested not just heavily in the RNA piece that is here in San Francisco, but we built an entire delivery group in West Point, PA. The thing that continues to differentiate Merck is that we have people with decades of experience in pharma R&D, drug safety, metabolism, pharmacokinetics. . .

Outside of RNA as a therapy in itself, he also talks about Merck's use of the technology to better understand its small-molecule targets. It's not something that you'll ever see press releases about, but trustworthy data of that sort is very useful and important. As the Xconomy interviewer notes, Wall Street values this sort of thing as basically zero (partly because you can't see the results of it for quite a while, if they're ever made public at all), but the value inside the company can be significant.

Of course, there can be things that happen inside drug companies that significantly destroy value, too, and it's not like the stock market can see (or understand) many of those, either, but that's a topic for another post entirely. . .and on a not perhaps unrelated note, one part of the interview above seems to suggest that "POS" is an internal Merck acronym for. . .wait for it. . ."probability of success". I, uh, kid you not.

Comments (6) + TrackBacks (0) | Category: Business and Markets | Drug Assays | Drug Development

January 26, 2010

The Infinitely Active Impurity

Email This Entry

Posted by Derek

Yesterday's post touched on something that all experienced drug discovery people have been through: the compound that works - until a new batch is made. Then it doesn't work so well. What to do?

You have a fork in the road here: one route is labeled "Blame the Assay" and the other one is "Blame the Compound". Neither can be ruled out at first, but the second alternative is easier to check out, thanks to modern analytical chemistry. A clean (or at least identical) LC/MS, a good NMR, even (gasp!) elemental analysis - all these can reassure you that the compound itself hasn't changed.

But sometimes it has. In my experience, the biggest mistake is to not fully characterize the original batch, particularly if it's a purchased compound, or if it comes from the dusty recesses of the archive. You really, really want to do an analytical check on these things. Labels can be mistaken, purity can be overestimated, compounds can decompose. I've seen all of these derail things. I believe I've mentioned a putative phosphatase inhibitor I worked on once, presented to me as a fine lead right out of the screening files. We resynthesized a batch of it, which promptly made the assay collapse. Despite having been told that the original compound had checked out just fine, I sent some out for elemental analysis, and marked some of the lesser-used boxes on the form while I was at it. This showed that the archive compound was, in fact, about a 1:1 zinc complex, for reasons that were lost in the mists of time, and that this (as you can imagine) did have a bit of an effect on the primary enzyme assay.

And I've seen plenty of things that have fallen apart on storage, and several commercial compounds that were clean as could be, but whose identity had no relation to what was on their labels (or their invoices for payment, dang it all). Always check, and always do that first. But what if you have, and the second lot doesn't work, and it appears to match the first in every way?

Personally, I say run the assay again, with whatever controls you can think of. I think at that point the chances of something odd happening there are greater than the chemical alternative, which is the dreaded Infinitely Active Impurity. Several times over the years, people have tried to convince me that even though some compound may look 99% clean, that all the activity is actually down there in the trace contaminants, and that if we just find it, we'll have something that'll be so potent that it'll make our heads spin. A successful conclusion to one of these snipe hunts is theoretically possible. But I have never witnessed one.

I'm willing to credit the flip side argument, the Infinitely Nasty Impurity, a bit more. It's easier to imagine something that would vigorously mess up an assay, although even then you generally need more than a trace. An equimolar amount of zinc will do. But an incredibly active compound, one that does just what you want, but in quantities so small that you've missed seeing it? Unlikely. Look for it, sure, but don't expect to find anything - and have 'em re-run that assay while you're looking.

Update: I meant to mention this, but a comment brings it up as well. One thing that may not show up so easily is a difference in the physical form of the compound, depending on how it's produced. This will mainly show up if you're (for example) dosing a suspension of powdered drug substance in an animal. A solution assay should cancel these things out (in vitro or in vivo), but you need to make sure that everything's really in solution. . .

Comments (33) + TrackBacks (0) | Category: Analytical Chemistry | Drug Assays | Life in the Drug Labs

January 25, 2010

GSK and Sirtris: A Bit More

Email This Entry

Posted by Derek

Nature has a short item on the Pfizer paper that questions the reproducibility of some key sirtuin work (covered here and here). There are some good points to temper the pessimism. Leonard Guarente of MIT, a key pioneer in the field, says:

". . . that the latest findings are neither surprising nor worrisome. The compounds may work only with fluorophore-conjugated peptides in vitro, says Guarente, but the situation is different in cells and in animals. The Nature paper, among others, went beyond the test tube and indicated that SIRT1 was more active in cells and in animals after application of the Sirtris compounds. Furthermore, resveratrol administration made no difference to the lifespan of yeast that did not have Sir23, indicating that the compound's action depends on this gene.

According to a statement from GlaxoSmithKline, Ahn's conclusion "ignores any possibility of direct activation of SIRT1 that may occur in a cellular environment that is not reproduced in vitro".

True, but there's still that problem of the Pfizer group not being able to reproduce the in vivo effects, which to me was perhaps the most worrisome part of the paper. Now, it's worth remembering that animal studies are not the easiest things in the world to do right, since there are so many variables. Small differences in animal strains and the like can sometimes throw things off severely. Even the Pfizer group admits this readily, with Kay Ahn telling Nature that "every in vivo experiment is a little bit different" and that "Under our conditions we didn't see beneficial effects, but we don't want to make a big conclusion out of those results."

That's an honorable way to put things, I have to say. Rather less honorable, though, at least to me, is David Sinclair's response from the Sirtris team. See what you think:

A possible explanation for the discrepancy, says Sinclair, is that Ahn and her colleagues did not provide information on the characterization of the compounds, which they synthesized themselves. So there is no way of knowing how pure they were or whether they're the same as those made by Sirtris. "The fact that mice died indicates that there may be an issue with purity,"

That's. . .not so good. In fact, it comes close to being insulting. Although I say a lot of uncomplimentary things about Pfizer's management, the fact remains that they have a lot of very good scientists there. And I assume that they can reproduce Sirtris's published procedures to make the sirtuin ligands. If they can't, frankly, that's Sirtris's fault. Everyone (well, everyone competent) checks out compounds thoroughly before putting them into an animal study. Asking "Are you sure you made the right stuff?" at this point is really a bit much, and doesn't do anything improve my opinion of Sirtris. (Which opinion actually was pretty good - until recently).

Comments (40) + TrackBacks (0) | Category: Aging and Lifespan | Drug Assays

January 12, 2010

The Sirtris Compounds: Worthless? Really?

Email This Entry

Posted by Derek

As followers of the drug industry know, GlaxoSmithKline famously paid $720 million to buy Sirtris Pharmaceuticals in 2008. Sirtris is the most high-profile shop working on sirtuins and resveratrol-like pharmacology, which subject has received a massive amount of press (some accurate, some scrambled). I've been following the story with interest, since the literature has me convinced that the aging process can indeed be modified in a number of model organisms, which makes me think that it could be in humans as well. And I also feel sure that advances in this area could lead to many profound medical, social, and economic effects. (GSK, though, is going after diabetes first with the Sirtris deal, I should add - among other reasons, the FDA has no regulatory framework whatsoever for an antigeronic, if I can coin a word.)

But whatever the state of the anti-aging field, doubts have crept in about the wisdom of the Sirtris purchase. Last fall, a group at Amgen published a study suggesting that some of the SIRT1/resveratrol connections might be due an an experimental artifact caused by a particular fluorescent peptide. Now a group at Pfizer has piled on in the Journal of Biological Chemistry. They're looking over resveratrol and a series of sirtuin activators described by the Sirtris group in Nature.

And unfortunately, they also find trouble due to fluorogenic peptides. The TAMRA fluorophore on their peptide substrates seems to pervert the assay. While the Sirtris compounds looked like activators initially, switching to the native peptide substrates showed them to be worthless. Further study (calorimetry) showed that the activator compounds bind to a complex of SIRT1 and the fluorescent peptide substrate, but not to SIRT1 itself (or in the presence of native substrate without the fluorogenic group). That's not good.

But worse is to come:

"Despite a lack of evidence for the Sirtris series of compounds as direct SIRT1 activators, we investigated whether the in vivo efficacy demonstrated by SRT1720 in several rodent models diabetes could be validated and attributed to indirect activation of SIRT1. We therefore attempted to reproduce the in vivo efficacy for SRT1720 in mouse models of type 2 diabetes previously shown. . ."

That word "attempted" should tell you what comes next. The reported high dose of the compound (100 mpk) resulted in weight effects and death. The reported low dose (30 mpk) showed no effects at all on any diabetic parameters, but instead seemed to lead to increased feeding and weight gain. To complete the debacle, the Pfizer group screened the Sirtris compounds through a broad panel of assays, and found that all of them hit a number of other targets (and appear significantly worse than resvertarol itself, which is no one's idea of a clean compound to start with).

Basically, these folks have thrown down the gauntlet: they claim that the reported Sirtris compounds do not do what they are claimed to do, neither in vitro nor in vivo, and are worthless as model compounds for anything in this area of study. So what is GSK going to have to say about this? And what, if this paper is at all accurate, did they buy with their $720 million?

Comments (125) + TrackBacks (0) | Category: Aging and Lifespan | Business and Markets | Drug Assays

January 5, 2010

Run It Past the Chemists

Email This Entry

Posted by Derek

I missed this paper when it came out back in October: "Reactome Array: Forging a Link Between Metabolome and Genome". I'd like to imagine that it was the ome-heavy title itself that drove me away, but I have to admit that I would have looked it over had I noticed it.

And I probably should have, because the paper has been under steady fire since it came out. It describes a method to metabolically profile a variety of cells though the use of a novel nanoparticle assay. The authors claim to have immobilized 1675 different biomolecules (representing common metabolites and intermediates) in such a way that enzymes recognizing any of them will set off a fluorescent dye signal. It's an ingenious and tricky method - in fact, so tricky that doubts set in quickly about the feasibility of doing it on 1675 widely varying molecular species.
And the chemistry shown in the paper's main scheme looks wonky, too, which is what I wish I'd noticed. Take a look - does it make sense to describe a positively charged nitrogen as a "weakly amine region", whatever that is? Have you ever seen a quaternary aminal quite like that one before? Does that cleavage look as if it would work? What happens to the indane component, anyway? Says the Science magazine blog:

In private chats and online postings, chemists began expressing skepticism about the reactome array as soon as the article describing it was published, noting several significant errors in the initial figure depicting its creation. Some also questioned how a relatively unknown group could have synthesized so many complex compounds. The dismay grew when supplementary online material providing further information on the synthesized compounds wasn’t available as soon as promised. “We failed to put it in on time. The data is quite voluminous,” says co-corresponding author Peter Golyshin of Bangor University in Wales, a microbiologist whose team provided bacterial samples analyzed by Ferrer’s lab.

Science is also coming under fire. “It was stunning no reviewer caught [the errors],” says Kiessling. Ferrer says the paper’s peer reviewers did not raise major questions about the chemical synthesis methods described; the journal’s executive editor, Monica Bradford, acknowledged that none of the paper’s primary reviewers was a synthetic organic chemist. “We do not have evidence of fraud or fabrication. We do have concerns about the inconsistencies and have asked the authors' institutions to try to sort all of this out by examining the original data and lab notes,” she says.

The magazine published an "expression of concern" before the Christmas break, saying that in response to questions the authors had provided synthetic details that "differ substantially" from the ones in the original manuscript. An investigation is underway, and I'll be very interested to see what comes of it.

Comments (46) + TrackBacks (0) | Category: Analytical Chemistry | Biological News | Drug Assays | The Scientific Literature

December 10, 2009

Selective Scaffolds

Email This Entry

Posted by Derek

We spend a lot of time in this business talking about molecular scaffolds - separate chemical cores that we elaborate into more advanced compounds. And there's no doubt that such things exist, but is part of the reason they exist just an outcome of the way chemical research is done? Some analysis in the past has suggested that chemical types get explored in a success-breeds-success fashion, so that the (over)representation of some scaffold might not mean that it has unique properties. It's just that it's done what's been asked of it, so people have stuck with it.

A new paper in J. Med. Chem. from a group in Bonn takes another look at this question. They're trying to see if the so-called "privileged substructures" really exist: chemotypes that have special selectivity for certain target classes. Digging through a public-domain database (BindingDB), they found about six thousand compounds with activity toward some 259 targets. Many of these compounds hit more than one target, as you'd expect, so there were about 18,000 interactions to work with.

Isolating structural scaffolds from the compound set and analyzing them for their selectivity showed some interesting trends. They divide the targets up into communities (kinases, serine proteases, and so on), and they definitely find community-selective scaffolds, which is certainly the experience of medicinal chemists. Inside these sets, various scaffolds also show tendencies for selectivity against individual members of the community. Digging through their supporting information, though, it appears that a good number of the most-selective scaffolds tend to come from the serine protease community (their number 3), with another big chunk coming from kinases (their number 1a). Strip those (and some adenosine receptor ligands and DPP inhibitors, numbers 11 and 8) out, and you've taken out all the really eye-catching selectivity numbers out of their supplementary table S5. So I'm not sure that they've identified as many hot structures as one might think.

Another problem I have, when I look at these structures, is that a great number of them look too large for any useful further development. That's just a function of the data this team had to start with, but this gets back to the question of "drug-like" versus "lead-like" structures. I have a feeling that too many of the compounds in the BindingDB set are in the former category, or even beyond, which skews things a bit. Looking at a publication on it from 2007, I get the impression that a majority of compounds in it have a molecular weight greater than 400, with a definite long tail toward the higher weights. What medicinal chemists would like, of course, i