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

« Sort Of Like A Wine Cellar | Main | Outside Reading »

December 27, 2005

What Makes an Ugly Molecule?

Email This Entry

Posted by Derek

A while back, I mentioned in passing that some chemical structures were ugly ones from a med-chem perspective. That prompted a reader to ask, very appropriately, what makes a structure ugly. The quick answer is easy, and a bit embarassing: an ugly drug structure is one that looks too different from other things that we already know are drugs.

Hah! There's the "Rule of Five" in a nutshell for you. But there's something to that approach, although I don't approve of it being used religiously. If a compound deviates wildly from the kinds of structures that we already know work, you do run some risks. Those deviations can be in sheer size (molecular weight), where we know from bitter experience that many synthetic compounds up in the high hundreds of daltons seem to have trouble being absorbed and transported around the body. Not all of them - but more and more as you go up.

Or you can deviate in things like the number of polar functional groups. Way too few of them, we also know, and your compound is so greasy that it also can have absorption problems, and the liver will tend to rip it to pieces even if it gets that far. Way too many polar groups, though, and the compounds seem to have problems getting through cell membranes. Again, these aren't written in stone, but your chances of success decrease the further away that you get and your research efforts should be planned with those odds in mind.

Beyond these parameters, there are specific chemical reasons to dislike a structure. If it has strongly reactive functional groups in it, there had better be a good reason for them. After all, there are a lot of things in the body to react with, almost all of which you'd want to avoid. Some chemotherapy agents fall into this category, which goes a long way toward explaining their low toxicity threshold. To give you an idea, a structure with an acid chloride in it will be dismissed out of hand, and an epoxide will be regarded as guilty until proven innocent.

Reactive groups can also include things that are chemically stable, but will fall apart under physiological conditions. Very acid-labile groups, for example, are probably not going to make it through the hydrochloric acid bath of the stomach (unless you coat the tablet to make it through into the intestine). And something like a simple methyl ester probably isn't going to survive, either, because there's an awful lot of acylase/hydrolase enzyme activity floating around in the intestines and in the blood. (You can use these effects to unmask your active drug once it's in the body - a "prodrug" - but that's another story).

And, finally, we reserve the right to call a structure ugly if it looks just plain too hard to make. After all, we generally have to make hundreds (sometimes thousands) of related analogs during the early development of a lead structure, and if the shortest route into the series is fifteen steps long, that's just not going to happen. The odds of success go down as the difficulty of the chemistry goes up, so all things considered, we'd rather work on something that we can deal with. There are plenty of compounds that are reasonable by all other criteria except this one, I'm sad to say, which means that we pay attention any time an interesting new synthetic method comes along. It might lead us into chemical spaces that no one's had a chance to explore (or claim!)

Comments (24) + TrackBacks (0) | Category: Drug Development


1. Paul J. on December 27, 2005 2:40 PM writes...


Posting on your vacation? For shame....

Thanks for the thoughts about the Rule Of Five.

I have added the rule of five to my ever growing list of "Silent Killers"( right after molecular modeling).

On a serious note, while the Rule o' Five is a reasonable litmus test for passive transport, is there anything for active transport? It seems that transport problems could be greatly reduced if the prodrug approach was used to actively shuttle a compound across the membrane/barrier of interest via active transport and was then clipped off. Plus, knowledge of active transporter SAR would allow more chemical space to be used for therapeutics, thereby increasing number of things that are considered "drug like".

Permalink to Comment

2. gil on December 27, 2005 11:10 PM writes...

There is no cure for arthritis. However, Celebrex can help reduce the pain, inflammation and stiffness of osteoarthritis and adult rheumatoid arthritis and ankylosing spondylitis.

Permalink to Comment

3. Petros on December 28, 2005 4:00 AM writes...

Getting Spam comments now ( (gil=

I too well remember the discussions about "paper toxicioogy" i.e. the chemists take on the desirabilitry of certain structures. A series of potential Michael acceptors I worked on got a lot of flak for the perceived risk.

Permalink to Comment

4. Jeff Bonwick on December 28, 2005 7:35 AM writes...

Derek, a few questions if you're game:

(1) What are some of the ugliest drugs that work anyway?

(2) What have we learned from the exceptions?

(3) Are there examples of important drugs, discovered long ago, that we'd fail to discover today because they fall outside these parameters?

(4) What can we do to recapture our lost ignorance?

On item (4) I'm thinking in particular of a technique in numerical analysis called simulated annealing. As the name suggests, it was inspired by the mettalurgical process in which a metal is repeatedly heated and cooled to increase its strength. The reason annealing works is that it improves the metal's crystal structure by shaking it out of local energy minima.

The mathematical equivalent goes like this: suppose you have some hard optimization problem, and you have a way of making incremental improvements to an existing solution, but you don't have a way of finding the global optimum. In simulating annealing you periodically inject varying amounts of entropy into the solution and then explore that solution space for some number of iterations, basically to see if you got lucky. If so, that becomes your new baseline; if not, you return to the original and throw a different wad of entropy at it.

You see where I'm headed. I'd think you could do the same thing with drug discovery. Starting with a given compound, occasionally do some crazy (and initially worse) functional group substitutions, then apply your usual tradecraft to the resulting molecule.

I apologize in advance for even entertaining the thought that this might be a useful observation. I've probably either just described something you already do every day, or that was long ago tried and shown to be useless. But hey, you never know.

Permalink to Comment

5. Kay on December 28, 2005 8:06 AM writes...

Is it possible that some med-chems hurt their companies by taking the "rules" too seriously? If the answer is 'yes,' then perhaps simple rules should be ignored until they are based on mechanism and have low false-pos and false-neg rates?

On a related subject, which contemporary fad is causing the most damage to NME productivity? [the 2005 NME list is just pitiful:]

Permalink to Comment

6. LNT on December 28, 2005 8:23 AM writes...

From my observations at the two companies I've worked for, Lipinski's "rules" are used primarily as a guideline for the synthetic chemists only. Once a compound is "in the vial", reguardless of what it "looks" like, it will be put through testing and die or flourish by it's own merits. I have yet to see a good compound get "axed" early just because it violates a couple of the Lipinski rules.

Chemists can only make a finite number of compounds for a given project, so our thinking generally is that our "odds" of success can be increased slightly if we keep most of the molecules that we make within Lipinski's guidelines.

I always enjoy looking at SAR tables that compare cellular efficacy to enzymatic potency for an intracellular target. It's interesting to see that Lipinski's rules often hold up quite nicely: very large molecules and very polar molecules seldom have good cellular efficacy which is exactly what Lipinski's rules would predict.

Permalink to Comment

7. NJBiologist on December 28, 2005 5:23 PM writes...

I think Jeff Bonwick and LNT are headed toward a question I'm really curious about: how often do the Lipinski rules give a false negative?

Permalink to Comment

8. DL on December 28, 2005 6:39 PM writes...


Medicinal Chemists tend to do the exact opposite. They suck the entropy out of molecules. It's too bad since there are real consequences.


Permalink to Comment

9. Psychic Chemist on December 28, 2005 10:17 PM writes...

Does anyone think that enzyme inhibition assays should be replaced with cellular assays as the primary assays in a drug optimization program (and then use the enzyme assay as the secondary assay).

I dont know what is the feasibility of doing high throughput screening assays using cells. I have heard of a few companies which claim to be able to do that efficiently.

Although, I still think that there is no substitute for an animal experiment as early as possible in an optimization program.

Permalink to Comment

10. LNT on December 29, 2005 9:02 AM writes...

Psychic Chemist: There are too many variables to consider when looking at cellular activity: enzymatic (target) activity, metabolism, and cell permiability being the main ones. The cellular activity that you see is a conglomerate of those three variables and therefore SAR becomes very difficult.

I was involved in a project for a few monthes where we used cellular activity as the primary screen to drive the SAR. We couldn't make heads or tails of our data and eventually gave up the project.

The only time (I am aware) that cellular efficacy is the primary screen is when the target is extracellular (ie a GPCR).

Permalink to Comment

11. Jasper on December 29, 2005 11:25 AM writes...

In my opinion, the main failing of drug discovery is not whether the primary HTS assay is in vitro inhibition assay or a whole cell assay, but the insistence on pharma to use the target-based genes-to-drugs approach. Frankly, the gene-based target identification approach cannot discriminate between cause and effect – you need solid biological validation for that. What LNT writes is exactly correct – whole cell and ultimately, whole organ and whole animal drug biology is messy and hard to interpret. But, in my estimation, that should be the very reason for using them because it surely doesn’t make sense to tweak and patch molecules together based on a target that has no chance of having a biologic effect. Psychic Chemist: one of the new paradigms being evaluate in pharma is the approach you suggest – to screen for biological effect first and then confirm the more interesting effectors through a target-based in vitro assay. That’s probably doomed for failure, too, unless one is willing to (a) substantially validate the target in the human population or in a really, really predictive animal model; or, (b) forgo target-based SAR in favor of biology-driven SAR in a really, really predictive cellular model. Whatever paradigm is chosen, it’s likely to remain messy and hard to interpret.

Permalink to Comment

12. Daniel Newby on December 29, 2005 5:27 PM writes...

Jasper says "Frankly, the gene-based target identification approach cannot discriminate between cause and effect--you need solid biological validation for that."

The biological validation increasingly comes from combining pathology and gene sequence data. With a sufficiently large study population, you pretty much get a list of which genes are important in which syndromes.

I think this will be revolutionary for CNS disorders, which are strongly heritable and strongly influenced by transmembrane proteins. Whole animal screening has mostly been a failure in discovering targeted drugs for disorders like schizophrenia and migraine, instead giving "dirty" drugs that merrily fiddle with a dozen pathways.

Permalink to Comment

13. weirdo on December 30, 2005 5:18 PM writes...

"a really, really predictive animal model".

Aye, there's the rub. Name one. They're pretty rare. How many times has cancer been cured in mice? How long have various companies been working on p38 (or MMPs, or ???) based on all that wonderful pre-clinical animal data? No, thanks. You can keep your "predictive animal models".

And what self-respecting small-molecule chemist cares about "highly validated" targets that get screened only to find "ugly molecules", the subject of Derek's original post. It's as big a waste of time to try to turn crap molecules into drugs against a "validated target" as it is to work on a crap target with beautiful looking molecules.

So let's all go try to find the 20th p38 inhibitor to put into the clinic (only to fail once again). But it's a "highly validated target" (thus likely to get funded)!!!

As least when you work on a novel target, no one can tell you're wrong, (at least until you're into Phase 2).

Permalink to Comment

14. Kevin on January 1, 2006 7:26 PM writes...

Hi Derek - would you or anybody else out there know of a good medicinal chemistry textbook? I am a molecular biologist starting to venture into chemical biology. I'd like to get a better understanding of medicinal chemistry but am not sure which textbooks are good. Thanks!

Permalink to Comment

15. PsychicChemist on January 1, 2006 8:18 PM writes...

LNT, My experience with whole cell assays is mostly from antibacterial projects. You can spend months doing beautiful med chem optimizing a scaffold against an enzyme or a protein synthesis reporter system. However, if the compounds do not get into the cell or are pumped out by efflux systems, then you have nothing useful at the end. We found it more meaningful to use the cellular assay as the primary screen. Some times the SAR did not make sense, but whatever data we generated from the cellular assay was a lot more useful.

You also highlighted enzymatic (target) activity, metabolism, and cell permeability as the main issues while looking at cellular activity. What are your thoughts on non-specific protein binding and how much of an effect that would have on the cellular activity of a molecule. I mean, there are lots of miscellaneous proteins (and some may be in fairly high concentration relative to your target enzyme) in the cell that might have a special liking for your molecule and how would you screen against that.

Permalink to Comment

16. daen on January 2, 2006 9:40 AM writes...

Kevin: You could try "An Introduction to Medicinal Chemistry" (Graham L. Patrick, OUP 2001)

Permalink to Comment

17. Milo on January 2, 2006 5:16 PM writes...

I really like "The Organic Chemistry of Drug Design and Drug Action" by Silverman. Nice book, easy to read and good examples.

Permalink to Comment

18. Kay on January 3, 2006 5:57 AM writes...

The above conversation assumes that a chemist can look at a molecule, make a decision, and have that decision prove out more often than chance alone. The med chem literature continues to collect reports that essentially say "Lipinski's stuff doesn't work for our compounds, so we count this (or that), and it kinda sorta works." There is a complete absence of prospective experimentation ... it's a good thing that this portion of drug RD is not regulated, because they would never stop laughing in White Oak. This part of med chem appears to closer to kindergarten science (Johnny, how many rotatable bonds do we have?) than high school science.

It's my view that (1) Lipinskiism could be an important explanation of why Pfizer folks can't ever get anything done and (2) that our decisions are rarely better than chance alone. Our small molecule enterprise is in a shameful state.

Permalink to Comment

19. NJbiologist on January 3, 2006 7:15 PM writes...

Kay--I'm having a hard time finding information on how often the Lipinski rules get it wrong. Can you suggest a few papers?

Permalink to Comment

20. Kay on January 4, 2006 8:48 AM writes...

NJbiologist: I have read at least six. Start with J. Med. Chem., 45, 2615, 2002 and 47, 6104, 2004. I think that the general assumption that kindergarten rules can effectively make decisions that are capital efficient is a substantial underestimation of the complexity of the paying species.

Permalink to Comment

21. weirdo on January 4, 2006 10:10 PM writes...

In fact, if you read Lipinski's paper and understand what it says, there are NO molecules for which "Lipinski gets it wrong".

Lipinski's paper does not say "obey all these rules or your molecule will have poor bioavailability".

It is a retrospective analysis of molecules that behaved well and those that did not. The paper makes the conclusion that if your molecule violates MORE THAN ONE of the "rules", the probability of good bioavailability is much lower than molecules that violate only one or none of the rules.

This has been bastardized -- by managers and sellers of simple screening libraries who probably never read the whole paper -- to "obey all the rules or your molecule will suck, obey the rules and your molecules will shine".

But that's not what the paper concludes. And the main tenet is clearly correct: if your molecule is too large, too lipophilic, or too lipophobic, it is less likely to be bioavailable. Not impossible, but less likely, than smaller molecules in a certain lipophilicity range. But staying within the guidelines is no guarantee of success.

Having seen the results of several thousand PK experiments, this conclusion is most certainly correct.

Permalink to Comment

22. NJbiologist on January 5, 2006 8:28 PM writes...

Thanks for the reading and the comments. I haven't been at this very long; I've only seen one molecule which broke Lipinski's rules get tested, and sure enough, the rats didn't think too much of it (I seem to remember staggeringly low plasma levels with three different routes of administration). The thoughtful, roughly probabilistic approach makes sense, although (I think Derek has touched on this before) it's always easier to just cite a reason to not make a molecule/run an experiment.

And a followup to Kay's earlier comment--I don't think chemistry is unique. I'm sure the folks at FDA would find plenty of amusing decisions in the biology (and later) parts of the pipeline.

Permalink to Comment

23. Kay on January 7, 2006 8:16 AM writes...

I agree completely with Weirdo regarding "bastardized." The problem is that many people believe these rules to be true, and they harm their companies in the process. This then extends to a harm of public health. If I were Lipinski, I would go on a 'pulling heads from dark places' tour.

Permalink to Comment

24. NJbiologist on January 7, 2006 10:59 AM writes...

Kay--I think I agree with you and Weirdo about people taking the rules too far, but I part company with Weirdo on absolving the authors of these papers of all responsibility. The presentation that would make it obvious to me that these rules can be overdone would be to show not only correct predictions but also to show false negatives (compounds not meeting the rules but having good bioavailability) and false positives (compounds meeting the rules but not having good bioavailability). Unless I'm missing something, Lipinski et al. and Veber et al. don't do this; Lu et al. come close with their second figure, but you have to take a pretty good look at it to get any quantitative sense of how many compounds' behavior isn't predicted by the rules.

Permalink to Comment


Email this entry to:

Your email address:

Message (optional):

A Last Summer Day Off
The Early FDA
Drug Repurposing
The Smallest Drugs
Life Is Too Short For Some Journal Feeds
A New Look at Phenotypic Screening
Small Molecules - Really, Really Small
InterMune Bought