There’s a new paper out on a topic that is of great interest for medicinal chemists: what sort of chemistry is it that we’re spending all our time doing? It’s a study of the literature from 1984 to 2014, analyzed by reaction type and other factors, and here’s the take-home: “. . .of the current most frequently used synthetic reactions, none were discovered within the last twenty years, and only two in the 1980’s and 1990’s. . .” Those two, as you have probably already guessed, are palladium-catalyzed couplings (Suzuki-Miyaura and Buchwald-Hartwig). (Here are some other analyses like this from the last few years, for comparison – that last one is a fifty-year retrospective, for example).
Here’s a post on this latest paper from Wavefunction, who calls the conclusions of the paper both depressing and embarrassing. I’m trying to decide if I agree with that. Overall, I would expect any such analysis (a wide number of choices over time between a large number of alternatives) to show power-law behavior, with a relatively small number of reactions dominating the list and a long tail of others stretching out behind them, and that’s pretty much what we have. So I’m not amazed at all that twenty or so reactions make the up the bulk of the med-chem literature – I’d be amazed if that weren’t the case. Best-seller lists are a good analogy; the great bulk of the sales are in the top ranks of any such list, and I’d expect a roundup of the biggest-selling books of the last thirty years to show a roughly similar distribution.
One difference, though, is that fashions in books can change more quickly and more easily than do favorite chemical reactions. Don’t tell an unpublished author this, but the barriers to entry are lower. And while there are infinitely many ways to write a novel, every single amide-bond-forming reaction makes an amide, and no one’s going to discover a newer and more fashionable amide compound class. That limitation on the number of bonding patterns in organic chemistry will emphasize the power-law aspects of such reaction counts even more.
What we do have, though, are occasional ways in which existing bonding arrangements suddenly become easier to realize, and there you have the palladium-catalyzed couplings. If further work in the field eventually brings us to efficient, reliable C-C bond formation between plain sp3 carbons especially chiral ones, which I will deem provisionally the “zambodium-catalyzed alkyl coupling reaction”, then you can expect it to shoot to the top of the charts as well, since such bonds are the basic currency of all organic chemistry. No one would find it depressing or embarrassing that the good ol’ ZCAC reaction was being used all over the place – it would have to be used all over the place. You’d have to be an eccentric weirdo and deliberately trying to avoid it.
Now, the current metal-catalyzed coupling reactions are not quite in that class, of course. And I don’t think that anyone doubts that there’s been a proliferation of aryl-(hetero)aryl bond formation since the early 1990s in medicinal chemistry. It’s a combination of several factors. First off, such bond formations were not easy to do before palladium-catalyzed coupling, and in fact, they seemed rather magical once they began to become popular. Second, they’re experimentally fairly easy – sure, optimizing them can certainly be a chore, but you can usually slap a coupling reaction together under pretty standard conditions and expect some sort of product to come out the other end in most cases. And third, the reaction produces structures that are pharmacologically favorable. Biphenyl-type structures were already known as pharmacophores before the Suzuki reaction made them easier to get to, and the general hinge-binding behavior of kinase inhibitors matched up well with both it and the aryl-amine couplings (those two, the target space and the reactions, helped make each other popular).
More on that second point, which is an overall influence on any such list of popular reactions. Remember, in medicinal chemistry, we spend a lot of time analoging, fruitlessly, into the void. This means that there is no way that a reaction can become a top-twenty favorite without being experimentally easy and adaptable to a wide range of substrates. That’s the very definition of one; no one should be surprised to see them there. What’s worth asking, though, is how the compounds we make are being determined by what reactions we find most easy and versatile, and whether this compound set is a decent fit for what we actually need to have to prosecute our med-chem targets.
There’s where the objections come in to all those Suzukis. You cannot, I feel sure, make your way forever into target space by stringing aryl groups together like links of sausages, but that’s what the reaction does, and its products are a larger proportion of the available compounds than they probably should be. Similarly, I don’t think that a library of every available aryl carboxylic acid condensed with every available aniline is going to be the answer to all questions, either (but as the example of DNA-encoded libraries shows us, if you have access to such a set, you might as well screen it!) So what kinds of compounds do you need to address the wide world of targets, and what kinds of reactions are needed to produce them?
This new paper makes a comparison to natural products, and that’s probably as good a way to look at the problem as any. Evolution has been exploring target space for a long time, with a range of chemistries at its disposal (although certainly with its own limitations as to reagents and conditions, a fact that’s not always appreciated). As this analysis shows, the sorts of products we make in med-chem aren’t a very good fit into the natural products universe, and similarly, the sorts of reactions we use to make them aren’t, either. One result might be to call for medicinal chemists to try to deliberately run more varied reactions, and I can endorse that. There certainly are reactions that are under-used, and when a new transformation shows up in the literature, its incumbent on medicinal chemists to give it a try and see how well it works. But remember, as long as “more varied” means “harder to run and less general”, that call is not going to get as far as you’d like. So another call would be for efforts to make natural-product-like space more chemically accessible: we need more robust reactions to get there.
And while we’re thinking about natural products, remember that the term usually refers to the more exotic substances that organisms produce. But if you look at the whole universe of biomolecules, then it starts to look just as boring and stuck-in-a-rut as medicinal chemistry. Sure, we use amide couplings a lot – but not as much as ribosomes do (and with only twenty amino acids, too). Amides, carboxylate esters, glycosides, phosphate esters – the great huge bulk of the chemistry of life involves manipulating just those linkages and a few others, over and over and over. It manages not be boring.