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

« Merck's Strategy vs. Pfizer's | Main | Not Letting Pfizer's UK Site Go to Waste? »

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


1. anchor on February 7, 2011 10:28 AM writes...

Derek: read that paper and also your take. And, as a medicinal chemist, am thinking of Republicans and Democrats fighting in Washington DC for the soul of America! The fight goes on and there will be no winner at the end.

Permalink to Comment

2. Virgil on February 7, 2011 11:18 AM writes...

+1 vote here for phenotypic screening, from someone who is engaged in exactly that, and can't understand why anyone would want to screen anything unless they have a viable endpoint that means something clinically.

Of course, the issue with phenotypic screening comes back to the same problems underlying all screening - you only get out what you put in. If your library is full of crap (as many libraries are), it doesn't matter how relevant your screen is!

Permalink to Comment

3. petros on February 7, 2011 12:23 PM writes...

Both approaches can work but it interests me that the proponents of DOS are two academics rather than med chemists with experiences of the pitfalls of drug discovery

In particular their assertion of the value of DOS wrt protein-protein interactions seems questionable whereras some fragment based approaches have proved promising in respect of such difficult targets. Contrast the rational approach of the Hamilton group to such targets.

Permalink to Comment

4. iridium on February 7, 2011 12:27 PM writes...

I am still not quite sure why they call it DOS and not just combinatorial chemistry...

Permalink to Comment

5. Ed on February 7, 2011 12:52 PM writes...

The DOS advocates have included a lovely diagram. Therefore it must be better. LOL.

Seriously, DOS only exists so that academics such as Spring can fluff up their publication records with pointless musings such as this and extract money from gullible funding agencies keen to support the next next big thing.

Permalink to Comment

6. Lester Freamon on February 7, 2011 1:27 PM writes...

The debate is obviously about phenotypic screening vs target-based, biochemical approaches, and it is pretty obvious that DOS can do the former, while fragment is vastly superior for the latter--we're supposed to learn that we should do a phenotypic DOS screen first, find the target, then make it amenable to a fragment approach. But I think that approach (and target-based screening in general) fails the "serendipity" test, in which you allow some black boxes to exist in which what we don't know can help us. The elimination of all risk in modern drug discovery backfires, IMHO. DOS leaves room for a bit of magic (which could be polypharmacology/multiple targets, hitting targets you had previously ruled out, unique binding modes, etc) that has been key to most of the great drugs in history.

Permalink to Comment

7. EDELRIS on February 7, 2011 1:30 PM writes...

Some of the DOS properties are potentially interesting:
Complexity in order to increase the target selectivity (...but not too complex. see Hahn's model of molecular complexity), chirality and 3-D shape for the same reasons, new rings and scaffolds possibly coming from NP or even fully "synthetic" (83% of rings found in NP are not commercially available according to Schneider).
It is equally true that very large molecules are poor starting points, especially when aiming at orally available candidates.
Not trying to sell our offer (I know several companies engaged along the same lines), we have tried to apply the DOS strategy to "medchem" molecules, and even to 3D fragments.
We propose on a non exclusive basis collections of "small" DOS compounds for hit-finding activities (MW More info upon request at

Permalink to Comment

8. barry on February 7, 2011 1:31 PM writes...

We have been prone to episodic messianic fevers before. QSAR was going to replace med. chem. No, Rational Design was going to replace med. chem. No, CombiChem was going to replace med. chem. No, QSAR by NMR...
when the fever cools, med. chemists have one more tool to use as the problem warrants.
If there's a real candidate to render med. chem. obsolete, it's the quarter-by-quarter game of Merger/Acquisition/firing that has now put thousands of researchers out of work.

Permalink to Comment

9. EDELRIS on February 7, 2011 1:33 PM writes...

We propose on a non exclusive basis collections of "small" DOS compounds for hit-finding activities (MW below 500) and DOS fragments (MW below 300). More info upon request at

Permalink to Comment

10. OldLabRat on February 7, 2011 1:54 PM writes...

Must have been a slow editorial day at Nature, time to create controversy.
I agree with Barry in comment 8: both are tools to answer a properly framed question; use as appropriate.

Permalink to Comment

11. partial agonist on February 7, 2011 2:12 PM writes...

I like phenotypic screening also, but if you are an academic go ahead and try to highlight that in an NIH proposal. When you do, the review committee will lambast you for conducting a fishing expedition and tell you to kindly come back when you know precisely the target you are after and why.

Fragment-based screening seems ideal for crystallizable targets or targets amenable to NMR, calorimetry, or other such precise direct interaction studies. That seems to leave out a lot of targets though.

I tend to think you can do DOS smarter, avoiding ridiculously undruglike library members. If so it's a nice complement to fragment screening.

Permalink to Comment

12. Muse on February 7, 2011 2:33 PM writes...

Phenotypic screening is also an interesting topic on its own. What is the probability that the unknown target of a compound identified by phenotypic screening is also present in man or can it be that the target is rat or yeast specific for example? Is it realistic to develop a compound without knowing its target (I know it is the way things were done in the past) or do we need a defined human target before testing in the clinic with all the associated costs? Proving that we have a responder species for our Tox studies may also be tricky.

Permalink to Comment

13. GSDecay on February 7, 2011 2:52 PM writes...

DOS and fragments are both valid approaches to finding biologically active molecules and should be employed appropriately but the twin editorials conveniently don't mention the use of a well built HTS set with good properties which will usually outperform DOS if used properly. You can advance fragments without x-ray structure - you just have to be smart.
If DOS and fragments have been on the scene about the same time then their utility should be judged on how many clinical candidates they have generated. Those guys at Vernalis, Astex et al seem to be doing OK but how many DOS-derived candidates are there in the clinic?

Permalink to Comment

14. Doug on February 7, 2011 2:59 PM writes...

Every DOS library I've ever looked at has been physicochemically hopeless: huge MW, massive clogP, full of dodgy functional groups. Sure, you'll get a 10 mM unvalidated hit against something and a paper in Angewandte, but then what?

Cambridge fiddles while Sandwich burns.

Permalink to Comment

15. Anonlot on February 7, 2011 3:01 PM writes...

It is quite sad to see that some reputable chemists are still working in their own silo. There is simply no single method or pathway or formula in drug discovery. They are part of the reason why management thinks that drug discovery could be easily outsourced. If drug discovery could be conducted by using fragment based method alone, a single shop in China could replace everyone in the States.

Permalink to Comment

16. CMCguy on February 7, 2011 3:10 PM writes...

#8 barry says it well that these are tools and as #10 OLR suggests should be implemented as most suitable to the situation.

In terms of the "fads" or "episodic messianic fevers" I often wonder how much all these various over hyped paradigm shifts have both been a real distraction to progress in drug discovery/med chem (low productivity vs costs) while at the same contributing to creating an Executive view that med chem is easily replaceable/disposable function.

Permalink to Comment

17. barry on February 7, 2011 3:24 PM writes...

one scenario for phenotypic screening is playing out for Hsp90. A Phenotypic (anti-proliferation) assay turned up geldanomycin. The first hypothesis--that it was a kinase (Met) inhibitor crashed and burned, but it was found to bind to and inhibit Hsp90 (which is needed for the post-translational maturation of Met). The idea that one could block such a target (it has a hundred known clients) without killing mammalian cells was startling. Still, small molecule inhibitors (structurally unrelated to geldanomycin) are now in human trials against cancer.

Permalink to Comment

18. Plutonium Grandad on February 7, 2011 4:07 PM writes...

I'll chime in on the #8 barry is right bandwagon. Barry is right.

Usually, only History (often distant History) can judge whether something is a paradigm shift or a fad.

Caveat: fads are way more common.

Permalink to Comment

19. MoMo on February 7, 2011 4:27 PM writes...

#15 has it right. DOS is nothing but Diels Alder and other ring forming reactions in the guise of "diversity", but hey, to each their own and milk the synthesis/investor cow while you can!

The average investor doesnt know feces from shinola!

Permalink to Comment

20. pharmadude on February 7, 2011 6:14 PM writes...

What a dissapointing article. The best that fragment screening can ever hope to accomplish is to provide a small fraction of the lead molecules that HTS currently does. I don't think even the fragment community argues otherwise? DOS on the otherhand is..well its just HTS. The whole DOS thing is completely overblown to the point that its a borderline scam.

Permalink to Comment

21. Anonymous on February 7, 2011 8:38 PM writes...

#8 "QSAR was going to replace med. chem. No, Rational Design was going to replace med. chem. No..."

Ummm. QSAR is a form of 'Rational Drug Design'. Rational Drug Design *IS* med chem.

Anything other than rational drug design is shooting in the dark. The Shot Gun approach.

Permalink to Comment

22. Anonymous on February 7, 2011 8:50 PM writes...

There is no substitute for rational drug design.

Be thankful. This is what keeps the monkey and the machine from taking the medicinal chemist's job.

Permalink to Comment

23. Plutonium Grandad on February 7, 2011 9:30 PM writes...

#22 anonymous:

We have a saying down here in the South:
If you give an infinite number of rednecks an infinite number of shotguns and an infinite number of road-signs, they will eventually blast out all the great works of the world in Braille.

Permalink to Comment

24. Jose on February 7, 2011 10:46 PM writes...

"I am still not quite sure why they call it DOS and not just combinatorial chemistry..."

'Cause then Stu couldn't continue to posit that he has invented something startlingly novel...

Permalink to Comment

25. Cellbio on February 8, 2011 12:55 AM writes...

Love phenotypic screening, and have done it. A Diversity library of 25K compounds screened against primary human cell measures (cytokine, chemokine production, integrin upreg, proliferation, multiple cell types), focused collections and a Med Chem effort against a single target thoroughly profiled in same assays. Have thought a lot about it, so will offer my views.

Virgil, could you share the endpoints that are clinically useful? I am quite sure cellular and animal measures of activity are largely disconnected from clinical utility. That said, I like phenotypic endpoints as an addition to target screening as another source of clinical leads, and validating the biology with chemical intervention rather than genetic manipulation is stronger for the purposes of drug discovery.

Muse, my preference is primary human cell screens as the starting point, could see using rodents, but would not invest in yeast screens personally. That largely solves the target-in-humans issue.

The tox species question and general concerns about safety are more difficult. Logically, I could convince most that our "known universe" about phenotypic compounds with unknown target would actually be largely the same as target. In both we would know what the compound does not do (Kv channels and all the PanLabs stuff). We would have all the standard clinical chemistries etc from tox species. Missing would be the knowledge of potency at a target and selectivity amongst the target family. However, if the target schema is transferred from molecular target to biological target, say inhibition of TNF release to make it up, then we can, and did, show selectivity and SAR. One can also show the same biological impact in rodents and other tox species thereby validating them and assessing potency. One can also probe pathways to indicate which pathways are and are not inhibited during the cellular response, further demonstrating specificity, implying biochemical specificity. One can, and we did, also show that such compounds were biologically more selective than known drugs and all the target based med chem efforts we profiled as well. However, the inability to cover all biology with our screens, or reach great certainty that tox would predict accurately human response to exposure, the company I was with then would not go forward, fearing they would see something they could have predicted if only they know the target. At this point, it did not help to point out the logic was flawed in that safety signals seen generally come from unappreciated pharmacology, either due to the limited knowledge about the target, or off-target effects of the compound that were not revealed in screening.

Screening a few thousand very similar compounds, all hitting "the target" also revealed that there is no single target pharmacology, and to think so reduces the complexity of small molecule drug discovery to neat hypotheses that blow up spectacularly in the clinic. This also led me to think that for phenotypic screens, I did not need a wildly diverse library, but really good assays to see meaningful differences in activity of related compounds, and that the best starting point might be useful drugs that could have unappreciated pharmacological impacts that support efficacy. And we saw such examples, evident only if one screens many knowns against a target, but lost if you pick one compound to be a standard for a target.

An interesting side note of screening a diversity library is that all your hits are singletons with no apparent SAR. A quick structure similarity search can help populate some related space, but many stand alone in space and it is not easy to get the resources to make mini-libraries for stuff with no intermediates. In sum, give me our best compounds, let biology speak, then let's talk about what we know about targets.

Another note, related to academics doing it, is that the hit rate is relatively high. At 10 uM, the hit rate was so high as to be useless. At 1 uM, the hit rate was acceptable. But the best scenario, if possible, is run many assays in parallel and ask for selectivity up front. WIth a hit rate of about 2% in a primary assay at 1 uM, asking for the compounds not to inhibit all biology, a good thing I thought, limited the number of hits another 10x. An interesting side note, the push for potency in oncology screens created in several cases, to compounds that inhibited all tested biology. Good way to impact cell growth, probably not a great "targeted therapy".

Last point, curve shape matters too. If there is a single target with a single binding site the curve will show an expected behavior. Or should I say the real data and not a fitted curve with fixed Hill slope and set top and bottom. Careful examination of the dose response tells a lot, and is often missing from screens that find activities in the 10-50 uM range. Phenotypic screens can't be done without a rigorous understanding of dose and effect. Someone should coin a new phrase...pharmacology has a nice ring to it.

Excuse my long post, but love the topic. This experience convinced me that all drug discovery is empirical. Target, no target, I don't care. Lovely signaling diagram that I can comprehend, well then it is not reflective of the cell's true complexity or the variation form one cell type to another. Screen the damn thing and then you will know.

Permalink to Comment

26. Pete on February 8, 2011 8:18 AM writes...

Fragment-based approaches are effectively a framework for structure based design and the output from a fragment screen could be used to direct Diversity Oriented Synthesis (DOC). One argument that could be made against DOC is that the resulting compounds are of too high molecular complexity. However, the optimum molecular complexity depends on the power of the assay. I would challenge Hajduk’s assertion that fragment libraries are more diverse than DOC libraries. I am also not sure that reference 3 provides ‘ample evidence that larger molecules are less likely than smaller ones to succeed as drugs in clinical trials’. Comparison of FBDD and DOC is complicated by the fact that the former is directed at specific targets while the latter is typically used to augment generic screening collections with little or no reference to specific targets.

Permalink to Comment

27. Lester Freamon on February 8, 2011 9:26 AM writes...

Do you have a blog? If not, please start one. Always original.

Permalink to Comment

28. Cellbio on February 8, 2011 9:39 AM writes...

Thanks Lester, was recently thinking about it, but didn't think I had skill any where near my benchmark, Derek. I appreciate your nice words.

Permalink to Comment

29. JB on February 8, 2011 11:28 AM writes...

Large compounds are not inherent to DOS libraries- it's entirely possible to apply standard filters to the designs and many libraries now exist that use such constraints. Just because early DOS libraries were large doesn't mean it's impossible to make drug-like DOS compounds.
The other thing to note when talking about macrocycles is that they tend to have physical properties better than what their MW suggests they would be. Nature Reviews Drug Discovery 7, 608-624

Permalink to Comment

30. Lego Chemist on February 8, 2011 1:37 PM writes...

If you go to Toys 'R Us, you can buy Lego's in two versions: You can get kits that will build just one thing (you know ahead of time exactly what you are trying to do), or you can buy boxes of assorted Lego's and build different things from scratch.

I like both.

Permalink to Comment

31. Anonymous BMS Researcher on February 8, 2011 10:05 PM writes...

I agree, @CellBio: please start a blog -- and post pointers to it here, I'd like to read it!

Permalink to Comment

32. Zack on February 10, 2011 10:44 AM writes...

If you read the paper that the DOS proponents cite as an example of a "protein-protein interaction" inhibitor, its pretty revealing. I've written more here:

Permalink to Comment

33. Claudio on May 7, 2011 12:37 PM writes...

@ Zack

you made me laugh!

Anyway, to respond to Derek final question, I know for certain that there are people around working on PPIs using the fragment approach. And it's working well.

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