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

« Tailfins and All | Main | As Thin As a Soap Bubble »

May 31, 2005

Modeling the Modelers

Email This Entry

Posted by Derek

Comes the question, in a comment to the last post: "How much value is added by computer simulations?" Arrr. That's not easy to answer. I'd say, in some cases quite a bit, but in the majority of cases, none at all. And whether the former make up for the latter is a mighty close call. I know that there are molecular modeling success stories out there, and I've been on a project that started off dramatically with a (subsequently validated) modeling prediction. But I've been on others where the modeling was a total waste of time and effort.

Those linger in my memory, and surely account for some of my sceptical feelings about modeling in drug discovery. More specifically, there are a some major intellectual mistakes that I've seen simulations lead people into. The first one is something I alluded to recently - the awful temptation to believe that if you've seen a model of your molecule docking into a model of your target protein, then you've seen your molecule docking into your target. You haven't, you know. You've just seen someone's best guess, and the odds are excellent that it's wrong. Make a few more analog compounds, and the wonderful model that explains it all is likely to take on an unhealthy spotted appearance.

Ah, but that refines the model further, you say. And so it does - until the next analog blows that one, too. Still more refined! We have to be getting close now! But you can go through cycle after cycle of this stuff, and that leads to another trap: running the project for the sake of refining the model. It's an easy one to fall into, and it can always be justified by imagining what you'd be able to do with a simulation that explained all your compounds simultaneously. But you're not going to get one of those. As far as I know, no one ever has. And while you're chasing it, you're likely as not wandering away from the real purpose of your project, which was to find a drug. Remember? Even a drug whose binding mode you don't really understand will do, you know.

And one more pitfall is the way that modeling can constrain your thinking. If you really believe you're seeing reality (the first trap, above), then you might start ruling out whole classes of potential compounds. After all, they don't fit the model - why make 'em? And that violates one of my laws of medicinal chemistry, which I think I need to assemble into a list of their own: never talk yourself out of making an easy compound. The number of drugs that have been found by tripping over them is much larger than the number that have been found by homing in on them mathematically. Until the modelers come up with some more convincing mojo, I'll stick with my stumbling style.

Comments (5) + TrackBacks (0) | Category: Drug Development


1. The Novice Chemist on May 31, 2005 9:57 PM writes...

How many drugs (new or old) that have made it to market have come from or been helped by either combinatorial chemistry or computational chemistry?

Permalink to Comment

2. John Johnson on June 1, 2005 8:33 AM writes...

My guess is that if these techniques are refined and used properly, they'll help a whole lot more. In the meantime, overinterpretation and wishful thinking, as noted, will dampen the benefits.

Permalink to Comment

3. PharmaChemist on June 1, 2005 8:54 AM writes...

In my experience in drug discovery the impact of computational chemistry is almost entirely related to a couple of factors: 1) the quality of the computational chemist and the willingness of the medicinal chemists to work with him or her and 2) the target class of interest and the ability of the analytical group to get X-ray structures of the target-ligand complex.

If an X-ray structure can be generated with a compound docked inside, computational work can be invaluable. That's why the target class is important. We won't be turning out any useful GPCR-ligand X-ray structures anytime soon. But kinases and nuclear receptors (NR's) can often be crystallized quite nicely with lignands in tow. And in fact, the programs that have benefitted the most in my experience from computational work have been kinase and NR targets, while the work on GPCR's has been largely unhelpful.

Permalink to Comment

4. David Govett on June 1, 2005 1:41 PM writes...

Part of the problem might be that each human is a unique physiological system, so until it becomes possible to accurately model individual physiologies, pharmaceuticals will be administered with fingers crossed.

Permalink to Comment

5. Tom Bartlett on June 2, 2005 4:47 PM writes...

I use modelling to "de-prioritize"--not to "talk myself out of" molecules. I think modelling has been pretty helpful to me, with the caveats of having a good modeller and a good Xtal structure to work from.

Permalink to Comment


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