I have a couple of nominees to get things rolling. For reagent, I would like to advance the montmorillonite clay stuff. I cannot count how many papers I have seen on its use as a Lewis acid, catalyst, and all-around good thing to have, but I have never used it myself, never spoken with anyone who has, and never (to my recollection) heard it suggested as a possible thing to try when someone encountered a synthetic problem. For all I know it's a fine reagent, but its footprint does not seem to be very large. I actually have used benzotriazole, but I've never seen an actual container of montmorillonite K-10.

For general technique, I'm tempted to nominate ionic liquids. Man, are there ever a lot of publications on those things, but again, I've never actually encountered them in actual practice. I have heard second-hand of people trying them, so I guess that counts for something, but it still seems to be disproportionate compared to the avalanche of literature citations for the things. The craze seems to have peaked, but still not a week goes by that I don't see a paper.

Nominations? As with the book recommendation post, I'll assemble things into master lists.

I'm interested in hearing about things in several categories - best introductions and overviews of the field (for people just starting out), as well as the best one-stop references for specific aspects of drug discovery (PK, toxicology, formulations, prodrugs, animal models, patent issues, etc.)

Feel free to add your suggestions in the comments, or e-mail them to me. I'll assemble the highest-recommended volumes into a master list and post that. Just in time for the holidays, y'know. . .

In some cases, that's because there are some drug-drug interactions, a problem the FDA has recently addressed. The proton pump inhibitors, especially, are metabolized through the CYP2C19 pathway. That's a problem, since that enzyme is needed to convert clopidogrel into its active form (Plavix, as it comes out of the pill, is a prodrug - its thiophene ring needs to get torn open). This sort of thing has been seen many times before - it's one of the many headaches that you can endure in drug development as you profile the metabolizing pathways for your drug candidate and compare them to the other compounds your patient population might be taking. There are some combinations that just will not work (several involving CYP3A4, which is often the first one you test for), and it looks like we can add Plavix/2C19 to the list.

But the population genetics of the 2C19 enzyme are rather heterogeneous. About a third of the patients taking Plavix have a less-active form of the enzyme to start with, and they might not respond as robustly to the drug. The FDA has emphasized this effect in its latest public health warning. That's an opportunity for Effient, since it doesn't go through that metabolic route.

The In Vivo people point out, though, that this story isn't being driven by the usual players. It's not the FDA that's pushed to find this out, and it's not even Eli Lilly. It's Medco and Aetna. They studied their insurance claims data to see if the numbers supported the proton pump inhibitor/Plavix interaction, found that they did, and publicized their findings - and that led to an actual observational trial from BMS and Sanofi, which confirmed the problem. Now Medco is going further, and is actually running its own observational study comparing Plavix and Effient. Their theory is that the efficacy that Lilly showed compared to Plavix was driven by the (deliberate, one assumes) inclusion of a high number of poor metabolizers.

Medco is getting ready for generic Plavix, and trying to keep its costs down by making the case that the drug will do the job just fine for most patients. They could, on the other hand, end up making the case for Effient in that poor-metabolizing third of the patients, which would also be interesting. Lilly would presumably settle for that, although they'd like even more of the market if they can get it, naturally.

And I have to say: I like this sort of thing. I like it a lot. This, to me, is how the system should work. Companies are pursuing their own competing interests, but in the end, we get a higher standard of care by finding out which drug really works for which patients. The motivation to do all this? Money, of course, earning it and saving it. This may sound crass, but I think that's a reliable, proven method to motivate people and companies, one that works even better than depending on their best impulses. You could even build an economic system around such effects, with some attention to channeling these impulses in ways that benefit the greatest number of people. Worth a try.

Update: oh yeah, Pharmalot returns, too!

When I look back on it, it's a very good thing that my graduate school curriculum only featured classes during the first year. Because I was trying to get away with more and more by doing less and less, and those two trend lines were heading toward an intersection. (Another example of that from my grad-school past can be found here). In the end, the chrome-plated jaws of destiny did not quite snap shut on my academic career, but it was a near thing. I can well recall being assigned problem sets in a course during my first year of grad school, with a strong probability of having to be called up to the board to work out a random one from the list in front of the professor and the class, and just not getting around to doing them.

So more than once, I'd be called upon to present a problem I hadn't actually bothered to look at. A classmate of mine, Bill, had a similar approach to his work, and he and I would sometimes end up side by side at the board, quietly saying things to each other like "You do any of these?" "Nope, me neither. This one look like the Eyring equation to you?" At the same time, I was ceasing to take notes in the class, finding that (for whatever reason) I wasn't getting much out of the lectures, and seemed to be doing OK by reading the material.

The professor involved noticed me sitting there without a notebook day after day, and called me in for a chat. "You seem to have ceased bringing any sort of writing implement to my lectures", he said. "I presume that there's some reason for that?" I stammered out some line about how I found that I was able to concentrate more on the material when I wasn't having to worry about getting it down on paper, and I could tell that he didn't buy that one for a minute. "I see. . ." he said slowly, and let me go. The next lecture (and you knew this sentence would start out that way), he was up at the board talking about More O'Ferrall plots or something of the sort, and in the middle of explaining one said ". . .then when you move into this quadrant the transition state is affected like so and does that look OK to you, Derek?"

Zzzzzip! Some home-security monitor circuit in my brain tripped, and I returned to reality with the unpleasant sensation of having been dropped into my seat from a helicopter. "Umm. . .no mistakes that I can see", I said, which was certainly true, and the professor gave me a narrow-eyed look. "Yes. . .no doubt".

So no, this couldn't have gone on in that style for too much longer, and it was with relief that I moved on to full-time lab work. But I still have little patience for lectures I find uninteresting. I'm just glad that no one's passing out exams afterwards. . .

That gave a large number of possible combinations - nearly a million, actually, and in most cases, no correlations showed up. But in about 7,000 examples, a drug matched some other ligand set to an interesting degree. On closer inspection, some of these off-target effects turned out to be already known (but had not been picked up during their initial searching using the MDDR database). Many others turned out to be trivial variations on other known structures.

But what was left over was a set of 3,832 predictions of meaningful off-target binding events. The authors took 184 of these out to review them carefully and see how well they held up. 42 of these turned out to be already confirmed in the primary literature, although not reported in any of the databases they'd used to construct the system - that result alone is enough to make one think that they might be on the right track here.

Of the remaining 142 correlations, 30 were experimentally feasible to check directly. Of these, 23 came back with inhibition constants less than 15 micromolar - not incredibly potent, but something to think about, and a lot better hit rate than one would expect by chance. Some of the hits were quite striking - for example, an old alpha-blocker, indoramin, showed a strong association for dopamine receptors, and turned out to be an 18 nM ligand for D4, which is better than it does on the alpha receptors themselves. In general, they uncovered a lot of new GPCR activities for older CNS drugs, which doesn't surprise me, given the polypharmacy that's often seen in that area.

But they found four examples of compounds that jumped into completely new target categories. Rescriptor (delavirdine), a reverse transcriptase inhibitor used against HIV, showed a strong score against histamine subtypes, and turned out to bind H4 at about five micromolar. That may not sound like much, but the drug's blood levels make that a realistic level to think about, and its side effects include a skin rash that's just what you might expect from such off-target binding.

There are some limitations. To their credit, the authors mention in detail a number of false positives that their method generated - equally compelling predictions of activities that just aren't there. This doesn't surprise me much - compounds can look quite similar to existing classes and not share their activity. I'm actually a bit surprised that their methods works as well as it does, and look forward to seeing refined versions of it.

To my mind, this would be an effort well worth some collaborative support by all the large drug companies. A better off-target prediction tool would be worth a great deal to the whole industry, and we might be able to provide a lot more useful data to refine the models used. Anyone want to step up?

Update: be sure to check out the comments section for other examples in this field, and a lively debate about which methods might work best. . .

And I've also added another category for chemistry and pharma database sites. There you'll find a quick way to access the copious piles of information from Drugbank, Emolecules, ChemSpider, PubChem, DailyMed, Druglib, and Clinicaltrials.gov.

As always, if I've left a blog (or your blog!) off the list, drop me an e-mail and let me know about it. If it's of potential interest to the readership here, on it goes.

I notice that as I write this I have a PyMOL window open on my desktop; I use the program regularly to look at protein structures. Si monumentum requiris, circumspice.

This latest trial was a small one, but people have been starved for data on Zetia ever since it took a surprising hit (in the ENHANCE trial) suggesting that it might not be very efficacious. There's an ongoing larger trial that should answer this question once and for all, but those numbers won't be showing up for another two years. For now, anything that can help clarify what's going on is of great interest to Merck, its investors, and to cardiologists and their patients.

And Matthew Herper at Forbes is right: these latest numbers are disastrous. The study (funded by Abbott) isn't the greatest piece of clinical research in the world - it didn't study nearly as many patients as it was designed to, since it was halted early. (Here it is in the NEJM). But it still shows Niaspan as clearly superior to Zetia, and it makes a person wonder if taking Zetia is basically an expensive way to take a possibly-inadequate dose of simvastatin. In a way, the relatively small size of the study actually helps it a bit - getting numbers that definitive without having to go to much larger sample sizes isn't so easy in cardiovascular trials, so the feeling is that there much be something here.

As Herper's article details, Merck is trying to spin this as a big win for their competition, not a big loss for their own drug. But that comes close to being logically impossible: cholesterol lowering, like many other therapeutic areas, is nearly a zero-sum game. If patients take Niaspan (or any other competing drug), they're not going to be taking Zetia. This one was certainly a victory for Abbott (and generic niacin, for those who can take it), but it was a loss for Merck as well.

The FDA's not coming out of all this looking very good, either:

"How is it possible for a drug to have $4 billion in sales without any evidence of benefit?" says Harlan Krumholz, a cardiologist at Yale University. He said that the small size of the two imaging studies mean they couldn't render a clear verdict on Zetia. "But they don't instill any confidence in it either. " Douglas Weaver, head of cardiology at the Henry Ford Hospital in Detroit says: "We've used Zetia without sufficient amounts of clinical data to support it. Using it may be right, it may be wrong, but we don't know right now."

But it's worth remembering that Zetia's mode of action made perfect sense, and that it really does lower cholesterol to what you'd think would be a very beneficial degree. But it probably has several other effects beyond simple LDL lowering, and just looking at that number is clearly (in hindsight) not enough of a clinical surrogate marker. As the study authors put it:

If viewed properly, this hypothesis-generating finding is not an indictment of the overall importance of reducing LDL cholesterol for the purpose of preventing cardiovascular events, as illustrated by therapies based on statins or nonstatins (e.g., bile acid sequestrants). Rather, this adverse relationship may be attributable to the net effect of ezetimibe, a drug with diverse actions, not all of which are measured through its effects on intestinal cholesterol absorption and LDL cholesterol level. Taken together with a preexisting concern regarding the clinical effectiveness of ezetimibe, our findings challenge the usefulness of LDL cholesterol reduction as a guaranteed surrogate of clinical efficacy, particularly reduction achieved through the use of novel clinical compounds.

But as I recall, statins themselves were first approved based largely on lowered LDL, with better outcome data only showing up later. In that case, the surrogate marker paid off, but not this time. What all this is telling us, then, is that we don't know nearly as much about cholesterol and cardiology as we thought we did. And if we don't understand that area well enough, after all these years and all this effort, what parts of medicine do we really understand?

I'm awaiting confirmation of that diagnosis, which is worrisome for all sorts of other reasons, but whatever the cause, this is a loss for synthetic chemisty. Prof. Fagnou had published many interesting and useful papers on catalysis of bond-forming reactions, an area that's been growing steadily in importance for years and shows no signs of faltering. We need all the smart, capable people we can get working on such things, and I'm very sorry that we've lost one. Condolences to his family, colleagues, and friends.

These are long-running questions, but over the last twenty years some lessons have been learned. A new paper in J. Med. Chem. emphasizes one of the biggest ones: if at all possible, run your assays with some sort of detergent in them.

Why would you do a thing like that? Compound aggregation. The last few years have seen a rapidly growing appreciation of this problem. Many small molecules will, under some conditions, clump together in solution and make a new species that has little or nothing to do with their individual members. These new aggregates can bind to protein surfaces, mess up fluorescent readouts, cause the target protein to stick to their surfaces instead, and cause all kinds of trouble. Adding detergent to the assay system cuts this down a great deal, and any compound that's a hit without detergent but loses activity with it should be viewed with strong suspicion.

The authors of this paper (from the NIH's Chemical Genomics Center and Brian Shoichet's lab at UCSF) were screening against the cysteine protease cruzain, a target for Chagas disease. They ran their whole library of compounds through under both detergent-free and detergent conditions and compared the results. In an earlier screening effort of this sort against beta-lactamase, nearly 95% of the hits (many of them rather weak) turned out to be aggregator compounds. This campaign showed similar numbers.

There were 15 times as many apparent hits in the detergent-free assay, for one thing. Some of these were apparently activating the enzyme, which is always a bit of an odd thing to explain, since inhibiting enzyme activity is a lot more likely. These activators almost completely disappeared under the detergent conditions, though. And even looking just at the inhibitors, 90% of the hit set in the detergent-free assay went away when detergent was added. (I should note that control cruzain inhibitors performed fine under both sets of assays, so it's not like the detergent itself was messing with the enzyme to any significant degree).

They point out another benefit to the detergent assay - it seems to improve the data by keeping the enzyme from sticking to the walls of the plastic tubes. That's a real problem which can kick your data around all over the place - I've encountered it myself, and heard a few horror stories over the years. But it's not something that's well appreciated outside of the people who set up assays for a living (and not always even among some of them).

So, let's get rid of those nasty aggegators, right? Not so fast. It turns out that some of the compounds that showed this problem during the earlier beta-lactamase work didn't cause a problem here, and vice versa. Even using different assays designed to detect aggregation alone gave varying results among sets of compounds. It appears that aggregation is quite sensitive to the specific assay conditions you're using, so trying to assemble a blacklist of aggregators is probably not going to work. You have to check things every time.

One other interesting point from this paper (and the previous one): curators of large screening collections spend a lot of time weeding out reactive compounds. They don't want things that will come in and react nonspecifically with labile groups on the target proteins, and that seems like a reasonable thing to do. But in these screens, the compounds with "hot" functional groups didn't have a particularly high hit rate. You'd expect a cysteine protease to be especially sensitive to this sort of thing, with that reactive thiol right in the active site, but not so. This ties in with the work from Benjamin Cravatt's group at Scripps, suggesting that even fairly reactive groups have a lot of constraints on them - they have to line up just right to form a covalent bond, and that just doesn't happen that often.

So perhaps we've all been worrying too much about reactive compounds, and not enough about the innocent-looking ones that clump up while we're not looking. Detergent is your friend!

What is unusual is the pattern found when comparing the internal reports with the published versions that showed up in the literature. The authors found that:

"More than half the clinical trials that we included in our analysis (11 of 20) were not published as full-length research articles. For 7 of the 9 trials that were published as full-length research articles, a statistically significant primary outcome was reported, and for more than half these trials, the outcome specified in the published report differed from the outcome originally described in the protocol. Three of the four trials with an unchanged primary outcome had statistically significant results for the protocol-specified primary outcome. Secondary outcomes also frequently differed between the protocol and the published report. Thus, trials with findings that were not statistically significant (P≥0.05) for the protocol-defined primary outcome, according to the internal documents, either were not published in full or were published with a changed primary outcome. . .all the changes that took place between what was specified in the protocol, what was known before publication (as presented in the internal company research reports), and what was reported to the public led to a more favorable presentation in the medical literature. . ."

The authors go on to point out that changing a primary outcome after you see the data is, in fact, a statistical sin (although that's not quite the phrase they use!) You really can't go around doing that, because you can end up chasing after random chance (and avoiding that is the whole point of running well-controlled trials). This does not cover Pfizer and Parke-Davis with glory, but it's worth noting that there's plenty of blame to go around when it comes to this practice:

"Our study is based on a relatively small number of trials undertaken to test a single drug manufactured by a single company and its successors. Furthermore, if a major purpose of the studies we examined was to promote off-label uses of gabapentin, the selective reporting we observed could be more extreme than that observed for studies conducted for other reasons. Previous studies in different settings have shown evidence of these same biases, however. Indeed, selective outcome reporting does not appear to be limited to studies funded by drug companies. Chan and colleagues examined published trials funded by the Canadian Institutes of Health Research and found that 40% of stated primary outcomes differed between the protocol and the published report. In addition, we cannot be certain that selective reporting was a decision made by employees of Pfizer and Parke-Davis, since the authors of the published reports included nonemployees. We did not systematically assess the methodologic quality of the included trials as described in the publications we examined. Previous research has indicated that quality scores are higher for trials conducted by the pharmaceutical industry than for trials conducted by not-for-profit entities, although reports from industry-sponsored trials have potentially distorted the scientific record because of other, less easily measured study factors."

That doesn't get the folks who conducted these gabapentin studies off the hook, although I should note that Pfizer disputes the conclusions of this article (as you'd certainly think that they would). And it's also worth noting that some of its authors have done work for the plaintiffs in suits against Pfizer over gabapentin (thus all the familiarity with the internal company documents, which came to light during discovery proceedings). But again, I don't see how that negates the paper's conclusions, and if Pfizer has any hard data that would do so, I think they should produce it with all speed.

And no, it's just a coincidence that this post involve Pfizer, after I've been going on about their merger business all week. Unfortunately, I think that they're probably not the only company that could be pointed at. But we in the industry shouldn't have things like this for others to uncover in the first place. Should we?

Speaking at the Reuters Health Summit on Wednesday, Kindler said Pfizer has melded and reshaped its research and development facilities within 20 days of buying Wyeth on October 15. With previous huge mergers, he said, that process had taken "literally years."

. . .Swift reorganization of the two companies' research operations stands in contrast to "the distractions, the disruptions and the delays that have plagued mergers of our company and others in the past," Kindler said.

(1) I think that, in the years to come, that people are most definitely going to need medicines. And by that, I mean new ones, because there are a lot of conditions out there that we can't treat very well. As the world gets (on the average) older and wealthier, this need will do nothing but increase. In many cases, pharmaceutical treatment is cheaper than waiting and having surgery or the like, so there's a large scale cost-saving aspect to this, too.

(2) I also think that many of these medicines are still going to be small molecules. Now, biological products can be very powerful, and can do things that we can't (as yet) do with small molecules - mind you, the reverse is true, too. And I think that biologics will gradually increase their share of the pharma world as we find out more about how to make and administer them. But it is very hard to beat an orally administered small molecule for convenience, cost, and patient compliance, and those are three very big factors.

(3) What we're witnessing now is a huge argument about how we're going to make those small molecule drugs, where we're going to make them, and who will do all those things. And it's driven by money, naturally. We don't have enough new products on the market, which means that we have to sell the ones we have like crazy (which leads to all sorts of other problems, legal and otherwise). At the same time, we're having to spend more and more money to try to get what drugs we can through the whole process. These trends appear unsustainable, especially when running at the same time.

(4) But as Herbert Stein used to say, if something can't go on, then it won't. Right now, the only way out that companies can see is to cut costs as hard as possible (and market as hard as possible). Those both bring in short-term results that you can point at. Long-term, well. . .probably not so good. But in that same long term, we're going to have to find better ways of discovering and developing drugs. If we can improve that process, the fix can come from that direction rather than from the budget-cutting one.

(5) And those improvements don't have to be incredible to make a big difference. We have a 90% failure rate in the clinic as it stands. If we could just work it to where we only lose 8 out of 10 drug candidates, that would double the number of new drugs coming to the market, which would cheer everyone up immensely.

(6) The questions are: can we improve R&D in time? Can we improve it with the resources we have? I think that the demand (and thus the potential rewards) is too great for a solution not to be found, if there's one out there. And we still know so little about what we do that I can't imagine that answers aren't out there somewhere. Who's going to find them? How long will it take? Where are they? I've no clue. But that looks like the way out to me.