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: Don't miss Derek Lowe's excellent commentary on drug discovery and the pharma industry in general at In the Pipeline

In the Pipeline

« Predicting Toxicology | Main | Organic Synthesis: A Dead End For Graduate Students? »

June 13, 2012

Live By The Bricks, Die By The Bricks

Email This Entry

Posted by Derek

I wanted to highlight a couple of recent examples from the literature to show what happens (all too often) when you start to optimize med-chem compounds. The earlier phases of a project tend to drive on potency and selectivity, and the usual way to get these things is to add more stuff to your structures. Then as you start to produce compounds that make it past those important cutoffs, your focus turns more to pharmacokinetics and metabolism, and sometimes you find you've made your life rather difficult. It's an old trap, and a well-known one, but that doesn't stop people from sticking a leg into it.

Take a look at these two structures from ACS Chemical Biology. The starting structure is a pretty generic-looking kinase inhibitor, and as the graphic to its left shows, it does indeed hit a whole slew of kinases. These authors extended the structure out to another loop of the their desired target, c-Src, and as you can see, they now have a much more selective compound.
But at such a price! Four more aromatic rings, including the dread biphenyl, and only one sp3 carbon in the lot. The compound now tips the scales at MW 555, and looks about as soluble as the Chrysler building. To be fair, this is an academic group, which mean that they're presumably after a tool compound. That's a phrase that's used to excuse a lot of sins, but in this case they do have cellular assay data, which means that despite this compound's properties, it's managing to do something. Update: see this comment from the author on this very point. Be warned, though, if you're in drug discovery and you follow this strategy. Adding four flat rings and running up the molecular weight might work for you, but most of the time it will only lead to trouble - pharmacokinetics, metabolic clearance, toxicity, formulation.

My second example is from a drug discovery group (Janssen). They report work on a series of gamma-secretase modulators (GSMs) for Alzheimer's. You can tell from the paper that they had quite a wild ride with these things - for one, the activity in their mouse model didn't seem to correlate at all with the concentration of the compounds in the brain. Looking at those structures, though, you have to think that trouble is lurking, and so it is.

"In all chemical classes, the high potency was accompanied by high lipophilicity (in general, cLogP >5) and a TPSA [topological polar surface area] below 75 Å, explaining the good brain penetration. However, the majority of compounds also suffered from hERG binding with IC50s below 1 μM, CyP inhibition and low solubility, particularly at pH = 7.4 (data not shown). These unfavorable ADME properties can likely be attributed to the combination of high lipophilicity and low TPSA.

That they can. By the time they got to that compound 44, some of these problems had been solved (hERG, CyP). But it's still a very hard-to-dose compound (they seem to have gone with a pretty aggressive suspension formulation) and it's still a greasy brick, despite its impressive in vivo activity. And that's my point. Working this way exposes you to one thing after another. Making greasy bricks often leads to potent in vitro assay numbers, but they're harder to get going in vivo. And if you get them to work in the animals, you often face PK and metabolic problems. And if you manage to work your way around those, you run a much higher risk of nonspecific toxicity. So guess what happened here? You have to go to the very end of the paper to find out:

As many of the GSMs described to date, the series detailed in this paper, including 44a, is suffering from suboptimal physicochemical properties: low solubility, high lipophilicity, and high aromaticity. For 44a, this has translated into signs of liver toxicity after dosing in dog at 20 mg/kg. Further optimization of the drug-like properties of this series is ongoing.

Back to the drawing board, in other words. I wish them luck, but I wonder how much of this structure is going to have to be ripped up and redone in order to get something cleaner?

Comments (39) + TrackBacks (0) | Category: Alzheimer's Disease | Cancer | Drug Development | Pharmacokinetics | Toxicology


1. RD on June 13, 2012 7:57 AM writes...

Funny you cited two targets I've actually worked on. What I think we have here is a failure to imaginate. In both cases, the compounds are flat and aromatic. But binding sites aren't necessarily the narrow slots into which we fit the skinny tabs that they first appear to be. Things move, sometimes up and down.
OTOH, it's hard to take risks when your job is on the line. You need to be very careful and conservative and it might be this caution that leads to compounds like these.

Permalink to Comment

2. Dr. Demented on June 13, 2012 8:18 AM writes...

" soluble as the Chrysler building." Love it! A new ranking on the (in)solubility scale perhaps, for things even less soluble than the dreaded 'brick dust.'

Permalink to Comment

3. partial agonist on June 13, 2012 8:58 AM writes...

When the modeling guy says "but we can build into this hydrophobic pocket, and a meta-biphenyl fits just great!"

You need to ask him/her what sp3 things fit OK, and try those too, or instead.

Permalink to Comment

4. CanChem on June 13, 2012 9:13 AM writes...

One project I was associated with had its lead structure be an aromatic tricycle, with very specific substitution (change anything and it was dead), and not surprisingly this stuff was brick dust. Similarly as above, we found that tossing specific aromatic groups off a side chain really boosted in vitro activity, but when it was dosed in rats, we had 8/8 deaths. Autopsies showed that it crashed out in beautiful crystals inside the rats, and killed them. Guess that's what you get when your IC50's are in the ug/mL, but so are your max solubilities...

Permalink to Comment

5. Ex-med chemist on June 13, 2012 9:15 AM writes...

Unfortunately I witnessed this kind of med chem nightmare play out time and time again. I blame Suzuki, Buchwald, Hartwig et al for their wonderful reactions that work beautifully in arrays.

I used to joke with teams that made flat molecules that we live in a 3D world not a flat one.

I also buried my head in my hands when a team were actioned to fix a tox issue only to take the only sp3 carbon out and say voila, next candidate.

Permalink to Comment

6. Rock on June 13, 2012 10:07 AM writes...

Nice post Derek. But you failed to mention that in addition to all the pitfalls you listed, even the in vitro numbers should be considered suspect when the solubility falls below levels of detection in high throughput assays which usually contain some DMSO. Compounds like that love to precipitate out, aggregate, and stick to anything it can find to avoid be solvated by water. This phenomenon may help explain the lack of in vitro/in vivo correlation.

Permalink to Comment

7. Matt Soellner on June 13, 2012 10:14 AM writes...

My lab is responsible for the selective c-Src inhibitor work and I thought I would make a few comments:

No attempts were made to obtain a compound with "druggable" properties. I'm not concerned with such properties and will leave making drugs to the drug companies. I will state, however, that the inhibitor is soluble up to ~125 uM in cell culture media and aqueous buffer.

While I am not concerned with making a drug, a selective compound enables hypotheses about c-Src to be evaluated. Thus, I am concerned with making probes that function in a cell. If you look at the thousands of citations that use PP2 as a "selective" Src inhibitor, it's pretty clear that a new tool is needed.

Furthermore, our studies show:

(a) Abl is a tumor suppressor in breast cancer
(b) selective Src inhibition is more effective in slowing cancer cell proliferation (and lower cell toxicity).

I'm happy with what we have learned (and continue to learn) about Src signaling in cancer using our selective inhibitor. None of it would be possible without our compound.

I hope the conclusions we find about Src will someday influence drug discovery, but it's not my lab's mission (or goal) to make compounds that can serve as drugs. For example, the drug companies are pushing dual Src/Abl inhibitors in breast cancer trials, however, our data clearly shows it's not ideal to inhibit Abl.

Permalink to Comment

8. Vader on June 13, 2012 10:20 AM writes...

"The compound now tips the scales at MW 555, and looks about as soluble as the Chrysler building."

Historically, Chrysler's problem has been solvency, not solubility.

Permalink to Comment

9. NoDrugsNoJobs on June 13, 2012 10:23 AM writes...

Thanks for the follow up comments Matt, I really like it when we hear from the researchers themselves - a two way exchange of info is great!

Permalink to Comment

10. Morten G on June 13, 2012 10:24 AM writes...

Flat and aromatic (and few rotatable bonds) are proxies for highly conjugated compounds. Drug molecules tend to have a bit that's highly conjugated, yes?

Permalink to Comment

11. WCA on June 13, 2012 10:36 AM writes...

This is the formulators problem, not mine.


Permalink to Comment

12. barry on June 13, 2012 10:56 AM writes...

First, I take issue with equating the tally of sp2 centers with flatness. Think of BiNaphth (or Gleevec).
Second recall that Lipinski studied the set of all drugs in the U.S. pharmacopeia, not just CNS drugs. His distribution of lipophilicities would have looked rather different if he had.
Third is the more important but perhaps subtler matter of distinguishing a tight-binding tool compound from a drug. It makes perfect sense to build out a known hinge-domain-binding-motif to get a potent, selective inhibitor of the kinase of the month. But to get from there to a drug, one then has to do the hard work of replacing that initial motif with something proprietary and soluble. Tools like CAVEAT help. And a x-ray co-crystal structure can save years of work. But it takes work and time. And time is money.

Permalink to Comment

13. billswift on June 13, 2012 11:16 AM writes...

I would think that the more selective a drug's action, the less toxic it would be. Is it that, other things being equal, a larger compound would have more metabolic products (some potentially toxic) as it breaks down?

Permalink to Comment

14. Rico on June 13, 2012 11:33 AM writes...

wrt comments from the author...the most likely reason companies are pushing a dual inhibitor is because they couldn't make a selective one in the first place...kudos to you for making a useful cellular tool - we need more of this sort of work in academia AND in industrial labs.

Permalink to Comment

15. milkshake on June 13, 2012 11:38 AM writes...

awful properties of other companies lead compounds are helping to keep me employed. We are developing biodegradable polymers that produce stable micellar IV formulations of insoluble drugs. For us, the greasier the active compound is the merrier. I remember one particularly terrible compound that we were formulating had a para-terphenyl core decorated with two carboxymethyls in para, para' positions...

Permalink to Comment

16. barry on June 13, 2012 11:58 AM writes...

re:#15 Milkshake
The (p)-terphenyl core announces a coumpound that came from fragment-based screening. Heroic formulations are the only hope for such a monster. It can't be fixed by med. chem.

Permalink to Comment

17. luysii on June 13, 2012 12:23 PM writes...

The compound also looks like a perfect DNA intercalator, even forgetting the near adenine embedded in the structure

Permalink to Comment

18. luysii on June 13, 2012 12:24 PM writes...

The compound also looks like a perfect DNA intercalator, even forgetting the near adenine embedded in the structure

Permalink to Comment

19. Garrett on June 13, 2012 1:02 PM writes...

I'm not a chemist, so might you explain what a 'tool compound' is, or is supposed to do? When I hear that, I think of indicator compounds for a test to check for the presence or absence of something. In that case, why would you ever check to see if it was useful as a drug? (I wonder if the patient is digesting carbohydrates - let's give them an IV of Lugol's Iodine solution and see if they turn brown or purple).

Permalink to Comment

20. SP3PO on June 13, 2012 1:19 PM writes...

#5 Ex-medchemist: "I blame Suzuki, Buchwald, Hartwig et al for their wonderful reactions that work beautifully in arrays."

I blame over reliance on UV detectors for monitioring reactions and characterizing products. Most of the people I work with think TLC stains are a quaint artifact of the dark ages. If they can't see it on their HPLC, they don't believe it exists. And what's more UV-active than flat, conjugated, aromatic molecules?

Permalink to Comment

21. Hap on June 13, 2012 1:42 PM writes...

A tool compound, I think, is a compound that inhibits one enzyme or a known (narrow) set of enzymes selectively. Administering it can thus selectively shut off one pathway of a biological process and thus be used to determine how the process works, or can be used for comparison to other putative inhibitors of the same enzyme to determine if they are selective for that enzyme, or if they are doing something else in addition to (or instead of) binding to the enzyme.

Permalink to Comment

22. partial agonist on June 13, 2012 1:50 PM writes...

#19: "tool compound" = A compound known to selectively act at a certain target, generally the first or one of the first to do so, and it can be used to validate the target.

Selectivity is the key, and hopefully some ability to use with in vitro systems, and ideally in vivo.

It helps answer the question: What does this target do? For example, if you have a identified a snazzy new enzyme that you suspect plays a role in elevating blood pressure, you want a tool compound, a selective inhibitor of the enzyme, to test the hypothesis.

The tool compound may be totally inappropriate as a drug lead (not patentable, too short half-life, poor solubility, etc.) but it may give you support for your hypothesis and send you off looking for more drug-like molecules.

I would imagine that most marketed drugs stood of the shoulders of one or more proof-of-concept tool compounds.

Permalink to Comment

23. Andrew Ryan on June 13, 2012 1:54 PM writes...

#19 Garrett: "I'm not a chemist, so might you explain what a 'tool compound' is, or is supposed to do?"

A tool compound is one that shows the desired activity in your screening assay, as an agonist or antagonist. You use it to validate and calibrate your assay, so that you have some known activity to compare the unknowns in your screening library to. It doesn't have to be a drug--it can be toxic (although not in your assay if you're using cells), poorly absorbed, difficult to synthesize, etc.

If you're running a screen to find some completely novel biological property that hasn't been found before (as I have recently), it can be pretty challenging to make sure your assay will actually find what you're looking for!

Permalink to Comment

24. Hap on June 13, 2012 1:59 PM writes...

What substructures in author 4 would make it liable to hERG inhibition?

Permalink to Comment

25. CMCguy on June 13, 2012 2:19 PM writes...

#5 I would not blame those who created good solid chemistry as much as those who take in applications to far. In Derek's post on transition to medchem a few days ago these prime reactions on target for "quickly", "easily", and "reproducibly" lesson that do not get modulated with the the #2 lesson. Array synthesis/Combichem is not necessarily poor approach that can be misused as a tool.

Once indeed more medchem/R view used to the "Formulators problem" and many drugs became successful due to innovation in delivery systems required to overcome poor physical characteristic build in with less than thoughtful design. There remains much ongoing work and opportunities but trust mindset will change for earlier considerations to reduce the hurdles for the D people.

Permalink to Comment

26. Ted on June 13, 2012 2:25 PM writes...

I spent a portion of my past working as a med. chemist in the GPCR arena. I eventually formulated a crude, but general understanding of how SAR in this arena worked.

It turns out that "GPCR" is the acronym for the "Grease Plus Charge Receptor" class.

In most cases, you can improve the potency numbers by sticking a single charged group onto your biggest, greasiest lead. Conversely, adding a big greasy glob or two to an otherwise modestly bioavailable structure could do wonders for the cell screen numbers.

I always love when the thing gets so big, and so unwieldy, that the only possible hope is to go the prodrug route! That's right, we're going all in! Stick something more on! Yay!

On the flipside, these things are job security to a process chemist.


Permalink to Comment

27. partial agonist on June 13, 2012 3:42 PM writes...

#24- just a guess, but maybe the imidazole?

A number of imidazole-containing compounds have hERG affinity, though a lot of them don't also....ketoconazole and other antifungals, ondansetron, H3-receptor inverse agonists, maybe others

These are things that to at least some extent block hERG but don't fall into the typical basic tertiary nitrogen center (often a piperidine) flanked by aromatic or hydrophobic groups.

Permalink to Comment

28. Hap on June 13, 2012 4:31 PM writes...

Curious Wavefunction posted on hERG receptor-related toxicity (fen-phen), but the summary he had of hERG-problematic functionality didn't seem to fit 4.

The imidazole would make sense based on your nutshell, since it's in both the lead and the optimized compound, and it's basic, and there's grease all over the place. I just didn't know what substructure should have flagged the compound as a hERG risk.

Permalink to Comment

29. lynn on June 13, 2012 7:43 PM writes...

And many of those "grease plus charge compounds" filling up industrial libraries end up having detergent activity that lyses gram positive bacteria. Easy to kill MRSA with them in vitro. But, in practice, they're usually too serum protein bound to have much in vivo antibacterial effect.
Some nice antibacterials, like fosfomycin have no UV absorption.

Permalink to Comment

30. TX raven on June 13, 2012 7:50 PM writes...

GPCR drug design has moved away from that paradigm, in part due to tox issues. Low nM affinity compounds are routinely obtained with neutral or weakly basic ligands.

Permalink to Comment

31. anonymous on June 14, 2012 5:19 AM writes...

"...including the dread biphenyl"

Can someone please provide details on Derek's comment above, with examples of the precise nature of the "dread factor" if possible??? Thanks!!!

Permalink to Comment

32. processchemist on June 14, 2012 5:44 AM writes...


from the little I know, byphenyls are oxydized by Ist pass metabolism to hydroxy byphenlys, with hemolytic properties.

Permalink to Comment

33. BeenThereDoneThat on June 14, 2012 6:34 AM writes...

Biphenyl groups are well-documented privileged structures and may have a plethora of off target activities.

Permalink to Comment

34. iridium on June 14, 2012 6:44 AM writes...


try to calculate the contribution to the LogP....

Permalink to Comment

35. Matthew K on June 14, 2012 7:28 AM writes...

IANAC (I am not a chemist) but this is one of the most interesting posts I've read on this blog in two years. It's a small peek into the world of designing compounds and solving biological problems with atom structures, which is pretty amazing if you think about it. I hope you can give a few more insights like this - although I enjoy lots of the other stuff as well, of course.
Anyway - thanks for writing, there are a lot of us reading and lurking out here.
- MK

Permalink to Comment

36. Fred on June 15, 2012 8:16 AM writes...

#19 "I'm not a chemist, so might you explain what a 'tool compound' is...."

A tool compound is a compound made by a tool.

#20 "I blame over reliance on UV detectors for monitioring reactions and characterizing products. Most of the people I work with think TLC stains are a quaint artifact of the dark ages."

Personally, I consider it poor technique not to get both. They are often complementary, and the TLC will also often help you pick chromatography conditions to start with.

Permalink to Comment

37. Anonymous on June 16, 2012 7:44 PM writes...

Derek show us all how it's done! Please share a link to your most recent manuscript where you led optimization of a new molecule with great drug-like properties. I seem only to find a paper from 2006 in BMCL, but am sure I am searching incorrectly as I only find six chemistry publications from Derek B Lowe?!?

Permalink to Comment

38. Aspirin on June 19, 2012 2:01 PM writes...

Anonymous, show us all how it's done! I seem to find...essentially no paper with your name.

Permalink to Comment

39. KissTheChemist on May 31, 2013 6:57 AM writes...

In case no-one pointed this out yet, in Med Chem you HAVE to gamble at some point, not knowing if it will pay off or not, whether you stick to Lipinski's rules or anything else. There are plenty molecules that made it market with properties that don't fit the rules.

Oh, and you REALLY want to go there? Behave.

Permalink to Comment


Remember Me?


Email this entry to:

Your email address:

Message (optional):

The Last Post
The GSK Layoffs Continue, By Proxy
The Move is Nigh
Another Alzheimer's IPO
Cutbacks at C&E News
Sanofi Pays to Get Back Into Oncology
An Irresponsible Statement About Curing Cancer
Oliver Sacks on Turning Back to Chemistry