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

« Ten Years After: The Genomics Frenzy | Main | The Hideous Numbers of Compounds »

January 20, 2009

Diversity-Oriented Synthesis: Oriented The Right Way?

Email This Entry

Posted by Derek

Ever hear of Diversity-Oriented Synthesis? It’s an odd bird. DOS tries to maximize the number of structures and scaffolds produced from a given synthetic scheme – to find the most efficient ways to populate the largest amount of chemical space. In a way, it’s the contrapositive of natural product synthesis, which focuses all its effort into producing one specific molecule at a time. I should add that DOS isn’t about producing mixtures; its goal is discrete compounds, but plenty of them, and all over the map. (Here's more background from David Spring at Cambridge).

The point of this is to increase the diversity of compounds libraries for biological screening. And that’s traditionally been the concern of the drug companies, but (as far as I can tell) there’s very little DOS going on inside the industry. All the publications in the field, at any rate, seem to come from academia. Companies certainly do care about the diversity of their screening libraries, but they don’t seem to be addressing the issue through the “maximum diversity in the fewest steps” philosophy.

There’s a recent paper in Ang. Chem. that will give you a good flavor of what’s going on in this area. A group led by Adam Nelson at Leeds has published an interesting approach that relies on olefin metathesis. An ingenious use of protecting groups and sequential metathesis reactions builds up a wide variety of structural backbones pretty quickly. (Another key feature is the use of fluorous tagging for purification, which will be the topic of another future post around here). Metathesis was certainly a good choice, since that gives you a chance to form a lot of carbon-carbon bonds in a lot of ways, all using basically the same reaction conditions. In just a few steps (around five or six) they ended up with about 80 quite different scaffolds.

Stuart Schreiber, an early advocate of DOS, wrote up a “News and Views” piece for Nature about this paper, and he makes the case this way:

” The resulting products differ from the compounds found in most small-molecule screening collections. Typically purchased from commercial vendors, the compounds in such collections frequently lack chirality and are structurally simple. This means that they can bind to only a small number of biological targets. The compounds in commercial libraries also tend to be structurally similar — their 'diversity' is limited to variations in appendages attached to a small number of common skeletons. This undesirable combination of properties means that, although enormous numbers of compounds (often more than a million) are frequently tested in screenings, at great expense, in the case of undruggable targets relatively few biologically active 'hits' are found. In principle, a smaller library of compounds that contains a more diverse range of molecular shapes, such as those made by Morton et al., would provide both more hits for less money, and hits for the more challenging biological targets.”

I see where Schreiber is coming from, but there are some details being overlooked here. One big point is that smaller compounds actually tend to hit more targets, just not with as much absolute potency: that's the whole idea behind fragment-based drug design. Larger, more complex molecules tend to be more selective, but when they happen to fit, they can fit very well indeed. You need a huge pile of them to have a chance of finding one of those, though. (I think that a happy medium would be a DOS approach to not-very-large compounds, but that doesn't give you that much room to maneuver).

Another point is that the key thing about the collections you can buy is that they often depend on just a few bond-forming reactions. You get an awful lot of amides, ureas, and sulfonamides, since by gosh, those sure can be cranked out. To me, that’s the first thing that makes the Leeds compounds stand out: none of these classic library-making transformations was exploited. Unfortunately, the other things that make the Leeds compounds stand out aren’t necessarily good. For one thing, there are no basic nitrogens in any of the structures. The paper lists a big class of azacycles, but in every case, the nitrogens are capped with nosyl groups, which completely wipe out their character. And while it’s true that you can get biological activity without nitrogen, you’ll get a lot more with it. A useful extension of the chemistry would be to use some sort of (update: more easily) removable group on the nitrogens, so that each scaffold could be unmasked at the end – that would give you the basic nitrogens back, and you could then make a few amides and the like off of them for good measure.

The compound set is also heavy on alkenes, which isn't surprising, given the metathesis chemistry. There's nothing wrong with those per se, but it would be worth taking all the scaffolds through a hydrogenation reaction to saturate the bonds, giving you another compound set. Alternatively, if you want to be a real buckaroo, take them through a Simmon-Smith reaction and turn them into cyclopropanes - that could be messy, but cyclopropanes are very much under-represented in compound libraries, compared to how many of them could potentially exist. A bigger problem is that one of the linking groups the Leeds team uses is a silyl ketal. That’s not the most chemically attractive group in the world, nor the most stable, and as a medicinal chemist I would have avoided it.

That brings up another point about well, the point of these libraries. Schreiber makes the pitch that if we're going to do chemical biology on the tougher interaction targets (protein-protein, protein-nucleic acid, and so on), then we're going to need all the chemical diversity we can get. That's hard to dispute! But a lot depends on whether these compounds are meant to be in vitro tools, or real leads for drug discovery. You can put up with silyl ketals (or worse) if the former, but not for the latter. (Many medicinal chemists would say that if you have some functional group that you're just going to have to remove, then don't put it in there in the first place).

And that's the gap between academia and industry on this approach, right there. The in vitro tools, used to discover pathways and interactions, are more the province of the university labs, and the drug leads are more the concern of industry. As it stands now, the drug company folks look at many of the DOS libraries and say "Hmm. . .sort of, but not quite". That's probably going to change, and if I had to guess, I'd say that one way into industrial practice might be through chemical vendors. There are a number of companies who make their livings by offering unique building block compounds to the drug industry - as DOS matures, these people may sense a commercial opportunity and move in.

Comments (50) + TrackBacks (0) | Category: Chemical News | Drug Assays


1. Nick K on January 20, 2009 9:26 AM writes...

This approach is all very well (and I liked the use of fluorous tags to remove excess reagent), but those scaffolds don't look drug-like to my eyes...

Permalink to Comment

2. McChemist on January 20, 2009 9:54 AM writes...

A useful extension of the chemistry would be to use some sort of removable group on the nitrogens

Wait, isn't that the whole point of using nosyl groups?

Permalink to Comment

3. Aaron on January 20, 2009 10:04 AM writes...

I am not a med chemist, I am 15 years the junior to Dr. Schreiber, and he is clearly a very bright and clever chemist. Here is the but - DOS, to me, is a great academic exercise, but if we assume there is a near infinite number of chemical structures, I do not see how making them 80 at a time is any better than rational approaches taken today. But - I could be way wrong.

Permalink to Comment

4. Schreiber_the_ charlatan on January 20, 2009 10:07 AM writes...

So, where is the evidence? Schreiber has been arguing for 10 years that DOS will enable the discovery of compounds against undruggable targets. So far, I haven't seen DOS yield a single real molecule that targets anything, let alone a protein-protein interaction. (By real molecule, I mean nanomolar, with in vitro evidence of binding to the putative target.) What's mysterious is why Schreiber is still granted a forum to carry on with his nonsense, in the absence of experimental data to support his claims. Or why funding agencies are still devoting millions to this stuff without it actually delivering useful compounds -- for example, the NIH Molecular Libraries Initiative, and before that the NCI Initiative for Chemical Genomics, and before that the Harvard ICCB, and lately the massive DOS program at the Broad....

Permalink to Comment

5. RandDChemist on January 20, 2009 10:08 AM writes...

The obvious issue here (pick a metaphor): how IS diversity defined? It's been awhile since I've had the opportunity to read diversity-focused literature, but I've not seen a reasonable definition thereof.

Permalink to Comment

6. Derek Lowe on January 20, 2009 10:10 AM writes...

McChemist, I haven't had as good an experience with nosyl as some, I guess. I don't care for the thiol-based deprotection, and sometimes I think you have to beat on them a bit more than you'd want. But it's true, that for N-sulfonyl, nosyl is about as removable as they come.

Permalink to Comment

7. Ty on January 20, 2009 10:37 AM writes...

Pre-emptive, chemistry-driven library syntheses, no matter how sophisticated the chemistry is, have a very long shot at finding any biologically useful entity, I am afraid, since you rely completely on luck. High-content cellular assay screening could increase the chance of finding something that may do something, but then you are left with a long and arduous task of target ID. If you add up all the fuzzy things along the way, you may end up not knowing what you are doing. Hmm.. Definitely a form of art, an academic endeavor.

If you want to claim usefulness in drug discovery, you should develop chemistry with a tangible purpose in the drug discovery context. For example, start with a fragment hit for a certain target, develop DOS or whatever that can quickly and easily fill up every possible spatial vectors emanating from the fragment, and show us that this approach got you the better lead more quickly. Then I'll buy it.

Permalink to Comment

8. Anonymous on January 20, 2009 12:09 PM writes...


I am not a chemist and not in this field. Why the vituperative attack? Why the "my definition is the only right definition" attitude? What do you mean by useful compound? Is a compound that clarifies a pathway or target in a cell based screen not a useful compound? If not, do you bother to read the scientific biological literature? Or only the chemistry literature, and the project you're working on?

You seem to carry a lot of emotional baggage, my friend!

Permalink to Comment

9. Artimisinin on January 20, 2009 12:18 PM writes...

Nice post.

My impression for DOS is that it is often times more chemistry driven. It cares more about publication type of ideas. Once compounds are there then claim their usefulness in drug discovery.

I agree with Derek that people like chemical vendors may be the solution for DOS. They are willing to think on what the pharmaceutical companies will buy instead of getting grants and publications.

Permalink to Comment

10. jt on January 20, 2009 12:39 PM writes...

What would be so terrible about using the scientific method, it has worked for many years very effectively. With all due respect to DOS, I bet none of these folks made any platinum compounds or lithium derivatives. Anyway, I have a number of compounds on the market to treat diseases, and every single one of them has a sulfonamide in it. So I love sulfonamides.

Permalink to Comment

11. Jose on January 20, 2009 12:44 PM writes...

Schreiber is an exceedingly, scarily bright guy, no doubt about it. That said, the hype around DOS and chemical biology is reminiscent of many trends of the past years.... look at the yearly page count of the J of CombiChem c. 1998 and today. I think there is a lesson there. Moreover, the entire concept of academic drug discovery is a pretty little pipe dream (Dennis Liotta excepted).

Permalink to Comment

12. Hap on January 20, 2009 1:42 PM writes...

I wonder how good the libraries are - getting a mess of crap, even if it's diverse crap, probably isn't going to help discover drugs, because people would like to know exactly what's in the pot. It would also be good to know whether the increased core diversity is more effective than the standard functionalization diversity, and whether libraries containing diverse ring systems reach targets that other fishing methods can't.

Permalink to Comment

13. Eco on January 20, 2009 4:54 PM writes...

"Diversity oriented synthesis?"

-Does that mean big Pharma actually has to hire US citizens in addition to Chinese???

-Sorry, good article. But I couldn't resist.

Permalink to Comment

14. JB on January 20, 2009 7:52 PM writes...

Hap- all the Broad DOS compounds must pass >75% purity by LCMS to enter the screening collection.

Permalink to Comment

15. cantab chem on January 20, 2009 10:02 PM writes...

I have seen DOS performed, and I do agree there is not one tool for all problems. However, being scientific, DOS has presented some real opportunities for drug discovery where others have failed. Are the solutions simple or perfect? Well, when is discovery simple?

DOS was first formulated in academia, propagated by many groups, and is now being proved and refined in the real world. There hasn't been one incarnation of DOS which has a perfect mix of true diversity, quality, and drug-likeness, nor will there be. I back JB's comment that high quality has been achieved, when it was important.

How far will it go? Let's see - data will come. But to rail as commentors against what some academics have pursued, flogging them with limits required in drug discovery, is a bit brutal. When is the same standard held to other chemistry research? Say, perhaps, natural product synthesis or catalytic methods?

Permalink to Comment

16. Dennis on January 20, 2009 10:40 PM writes...

"My impression for DOS is that it is often times more chemistry driven. It cares more about publication type of ideas. Once compounds are there then claim their usefulness in drug discovery."

Which is never true for target oriented syntheses.

Permalink to Comment

17. Pipe dream? on January 20, 2009 10:59 PM writes...

Academic drug discovery is a pipe dream? Maybe you should talk to Richard Silverman at Northwestern or Ted Taylor at Princeton or Robert Holton at Florida State. New chemistry buildings are just one of the benefits. I am not sure if finding a new ligand to get 80% ee in the addition of dimethylzinc to benzaldehyde will build a new chemistry building.

Permalink to Comment

18. Chem 201 on January 21, 2009 12:07 AM writes...

I think it is too earlier to evaluate DOS as a pipeline for therapeutics. It seems that positive results from DOS so far have been in the development of molecular probes (such as Schreiber's HDAC inhibitor tubacin and Jared Shaw's HOXA13 hit).

Let's hope that with all the public money being poured into DOS that more positive results will come.

Permalink to Comment

19. Jose on January 21, 2009 2:12 AM writes...

Taxol doesn't count in the least; great chemistry, but the NIH and a zillion other labs contributed; a semi-synthesis is pretty far removed from discovery. Pemetrexed is a third tier oncologic with ~$200 million per annum in sales. Any sane marketing group would not proceed with development. Pregabalin- obviously counts, but such a bizarre little structure, and for such a black-box CNS area? Most pharma labs wouldn't touch that for good reason.
So.. all in, we have a scant handful over the past how many decades of how many groups with that stated goal? Colour me not convinced.

Permalink to Comment

20. Morten G on January 21, 2009 5:21 AM writes...

Nick K:

Isn't "they don't look drug-like to me" a circular argument if we are talking about structures that are very different from what's in the libraries available today?
If you were saying that they looked metabolically unsound then I could follow you but isn't the point of this DOS stuff that it is supposed to give you scaffold unlike what you would expect?

Permalink to Comment

21. Nick K on January 21, 2009 6:46 AM writes...

Morten G:
Point taken, but why do you think drugs are structurally so very much of a muchness? Is it because med chemists have traditionally made and tested heterocycles, or because heterocycles, for what ever reason, confer biological activity? I incline to the latter view.

Permalink to Comment

22. Ty on January 21, 2009 10:04 AM writes...

Interesting, 'diverse' thread.

No proof until one shows it, but I tend to agree with Nick #21. Why are there so many heterocycles in drug? Hydrogen bonding in the hydrophobic pocket constitutes the best binding motif in protein-ligand interactions. Conventional 'polar' groups (amide, hydroxy, etc), while good for H-bonding, pay a big penalty of desolvation getting into that pocket. Heterocycles are hydrophobic enough (little desolvation penalty) yet good enough for H-bonding. It's my 2 cents.

Permalink to Comment

23. eugene on January 21, 2009 10:29 AM writes...

"Academic drug discovery is a pipe dream? Maybe you should talk to Richard Silverman at Northwestern or Ted Taylor at Princeton or Robert Holton at Florida State. New chemistry buildings are just one of the benefits."

Wow, that's a great benefit. New chemistry buildings are really nice. You don't realize it until you've worked in both a new and an old one. Another great benefit is that all the drug discoverers are very poorly paid, 2 or 4 year limited contract, and working 60-80 hour weeks, while the boss gets all the fame and cash. And since academic drug discovery is doing so well, there is no reason to have all those drug companies. So, until there is permanent (long term contract), well-paid research staff at universities that have 40-50 hour work-weeks, Schreiber can fail. He can fail big and he can fail hard.

Thankfully, with his DOS failure so far, he has not managed to make industrial jobs for which his students are training, obsolete. Don't get me wrong, he's a great guy. He's just a boss in a terrible system. He should start a company and pay his graduates a living wage.

Academia being seen as the only place where good discovery takes place would be awful for chemistry as a profession the way it's set up today. It would depend on training too many people for a job that they can only do back at the university. Which, I suppose, is what some people are saying the problem is today already.

Permalink to Comment

24. Hap on January 21, 2009 11:23 AM writes...

75% purity doesn't impress me all that much - while it seemed better than I could make sometimes, it means that almost 25% of your sample could be something else, which seems to leave an awful lot of room for an impurity to have a significant biological effect. What level of purity is standard for compounds (not yet drugs) tested for activity at drug companies? That would seem to be the relevant standard for library purity.

Permalink to Comment

25. dave on January 21, 2009 12:19 PM writes...

nice post. this is a very typical academic application but not very useful in a practical, pharmaceutical, industrial setting. the fact that most small molecule drugs on the market are simple compared to natural products should be a hint--why make something more complicated than it needs to be. the scale up of any hit coming out of these DOS libraries is not likely to be commercially practicle and without LC/MS or HPLC purification of the librairies, the data is likely to be useless (75%???).

Also, in reality, how relevant is diversity? the diversity that Schreiber describes is left over ideas from the combi-chem/molecular diversity days which has not gone anywhere.

Permalink to Comment

26. Jose on January 21, 2009 12:41 PM writes...

I've personally been involved in three different wild goose chases for the true active component of a screening hit. One was a 2-5% impurity, one was a oxidation-rearrangement product, and the last was an aggregation will-o-wisp. These were all nominally >90% purities. 75% with scaffold diversity leaves an ocean of possibilities. Most likely outcome of this diversity is buckets of highly frustrating spectra....

Permalink to Comment

27. Hap on January 21, 2009 12:51 PM writes...


Someone with the ability to develop drugs would be an improvement, whether at universities or in companies. With the push in schools for patent income, if a professor comes up with a method, the likely outcome would be a professor-created startup leasing the patents from the university. The university and the professor will get paid if the idea works (and the university might get paid anyway). The employees (likely ex-students) will get paid a living wage but it'll be like grad school forever. An unpleasant thought, though it's better than wondering which bridges provide the best shelter and from what sources running water can be drunk without requiring medical attention.

Permalink to Comment

28. Ed on January 21, 2009 12:56 PM writes...

Eugene #23 - Schreiber has started another company - Forma Therapeutics. They aren't even limiting themselves to DOS chemistry!

Maybe they will even pay a living wage!

Permalink to Comment

29. MTK on January 21, 2009 12:58 PM writes...

The 75% number is just what JB(comment #14) said was necessary for submission at the Broad Institute and did not reflect what was done in the paper. I don't know if 75% is still the case there, because that sounds awfully low by today's standards. For example, for submission to NIH the requirement is now 90% purity as determined by LC at 230 nm.

I checked the supplemental data for the Nelson paper, they did not report the purity for all of the compounds made and the ones they did were determined by 500 MHz 1H NMR. As far as I can tell they did not conduct a final HPLC to purify these (I could have missed it though, since there are 537 pages in supp. info), so the purities are not that high, ranging from 50-90%.

From a practical standpoint that's not so bad actually. You generally do mass-directed LC on everything and if you go in with things that pure, you should have no problem meeting the 90% threshold.

Permalink to Comment

30. Hap on January 21, 2009 1:38 PM writes...

My other problem is the use of fluorous chemistry to make the libraries - it's a neat idea, but the reagents are more than a little pricey, and I don't know whether Curran's company would want a piece of the revenues from the use of the technology or would simply charge a lot at the front end to use it. There are alternatives that might work (one group used big flat aromatic tags and carbon as an absorbent), but I would assume that alternative methods would be encumbered by patents as well. (Even when the patents do expire, I can't imagine that the fluorous tags are going to be cheap anyway.)

Permalink to Comment

31. CMC Guy on January 21, 2009 1:56 PM writes...

If one is doing screening or even animal models a "75% purity" should be acceptable considering the variability in the biosystems, although overall confidence does increase if start with better material. Maybe it's not this way as much anymore since now Med chemists commonly do know what an HPLC does but I have run into a fair number of lead compounds "purified by chromatography with NMR >95% pure" that were

In terms of DOS it appears a good concept as thinking seems to be attempted exploration in 3D mode which can be overlooked in libraries/collections. Problem, which think Derek elaborates, is that diversity in chemical space is huge whereas biological systems can be intolerant demanding selectivity for the desired effects. It is usually the wisdom of old hand chemists/biologists that can drive to the necessary intersection.

Permalink to Comment

32. Marvin S. Yu on January 21, 2009 3:47 PM writes...


I work for Fluorous Technologies, Inc. and think that I can answer your questions.

First let me say that I am a regular reader of In The Pipeline and I was going to wait to answer your questions, since Derek said that fluorous would be the subject of a future post. Since you brought it up, however, I might as well chime in now.

The IP is not an issue if you buy the reagents from us or one of our distributors, such as Aldrich, and as long as you use the reagents and separation media for internal research purposes. There is an implied limited license granted upon purchase. If you were using the technology to make libraries to sell to third parties, however, that would require a explicit license from us. We certainly are not looking for any reach-through on compounds or drugs found using our products.

The cost of the reagents and separation media is considered high by many, but we like to think in terms of value. By supplying tools which facilitate a fast, simple, and general (key to the DOS libraries) purification method, we believe the technology saves money and time in the long term.

One piece of evidence to support this is some of the comments we received when applying a NIH Pilot Scale Library grant. We were awarded the grant, but the summary review included a statement that the study group believed we were being overly optimistic about how many compounds we could deliver given the budget and resources that we proposed. By the end of Year 1, however, we had met those targets in terms of number and type of compounds and actually came in slightly under budget.

Could we have done just as well not using fluorous techniques and using lots of automated chromatography equipment instead? We don't know. The grant funds us to make compounds, not compare methods so direct comparisons have not been done.

One group that has published a direct comparison in a library synthesis is the Fustero group from Valencia. They preferred the fluorous approach over others. Of course, that's just one example. In the end whether or not the value of a fluorous approach is worth the cost probably depends on a multitude of factors.

This is getting long, so I'll just wrap it up by saying that we do have a very dedicated group of repeat customers who clearly believe the value is there.

Permalink to Comment

33. Hap on January 21, 2009 4:21 PM writes...

I didn't mean to insult your tech or to say it was overpriced - the idea of fluorous chemistry is neat. I thought some of the material was expensive, but for lead compounds (for finding them and knowing what they are) it makes sense. I don't know what would work if one of the leads becomes a drug.

I also thought that Materia (the metathesis people) generally wanted a piece of revenues instead of (or in addition to) a fixed upfront cost, and I didn't know if others did similarly.

Permalink to Comment

34. Marvin S. Yu on January 21, 2009 4:37 PM writes...


I didn't think your comment was insulting at all. Those were legitimate questions which I hoped to have answered.

Permalink to Comment

35. J. on January 21, 2009 4:47 PM writes...

XTL use DOS chemistry for their HCV program and have a number of intersting patents filed/published for potent HCV inhibitors: They are clearly not from the Lipinski school of thought.

Permalink to Comment

36. J. on January 21, 2009 4:52 PM writes...

...worth pointing out of course that XTL sold their DOS program to Presidio for $4 Million up front and a potential further $104 Million in milestones.

Permalink to Comment