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.