With all modern advances and super-sexy scientific tools we currently have, randomness is still arguably the best source of truly innovative discoveries. And scientists do acknowledge that, so various “accelerated serendipity” techniques are flourishing in drug discovery, organic chemistry, chemical biology, and other fields.
Recent paper in Angewandte from Adam Nelson group at Leeds, describes one such approach, activity-directed synthesis (ADS). At first glance, the name and TOC seemed to point in the direction of in situ click-chemistry, pioneered by Barry Sharpless. But in fact it is more like an elaboration on the seminal “accelerated serendipity” paper from MacMillan lab.
The core idea of the method is multi-dimensional screening of conditions (solvent, catalyst, substrate, co-substrate or whatever else you can vary in the 96-well plate format) for identification of something new. MacMillan lab was looking for new reactions that give a reasonable yield. Warriner and Nelson are looking for new androgen receptor agonists. They streamlined the procedure by iterative improvement of the biological activity in TR-FRET assay. The big goal behind this is to mimic evolution of biochemical pathways, so that in the end the team will have both active molecule and a synthetic method for it.
The chemical workhorse for Nelson’s ADS is the promiscuous reactivity of rhodium carbenoids, which are generated from corresponding diazo precursors. Under different conditions and with different co-substrates they form mixtures of products. If some of these products have reasonable bioactivity, the corresponding conditions proceed to the next round, where the procedure is repeated with more focused set of co-substrates and catalysts.
In comparison to their previous report, this time the authors substantially expanded the range of tested derivatives. But this didn’t result in more potent compounds. And one can argue that it’s not surprising. As any evolution-based process, ADS is prone to flaws. If at an early branching point the selected path is not absolutely optimal, all subsequent generations will inherit this ‘defect’. Also there are two obvious biases in the whole approach: the first is the ‘yield bias’ (compound should be synthesized in appreciable quantity in the first place), the second is the ‘chemistry bias’ (obviously, chosen rhodium carbenoid chemistry limits the possible outcomes).
On the positive side, the method is a useful tool for exploring SAR (along with reaction scope, which should be kept in mind) and generating good lead molecule for further improvement. As a bonus, from this work Nelson group also discovered first enantioselective O−H insertion of rhodium carbenoid. Let’s see how they are going to further develop the approach.