The title of today’s post is reminiscent of Stallone’s old movie, but I want to talk about cliffhanging of a different kind. A recent paper in J. Med. Chem. by Bajorath is a follow-up to his previous review in that journal, covering the so-called activity cliffs in drug discovery. This concept relates to pairs of structurally similar compounds with a significant difference in potency. There are many kinds of activity cliffs (hydrogen bond, stereochemical, lipophilic), and they come from many different target classes. Here is a typical example:

http://pubs.acs.org/doi/abs/10.1021/jm401120g

As you can imagine, some target classes are notorious for their association with scaffolds that lead to cliffs. The presence of more than one activity cliff in a given series might complicate compound optimization and turn it into a complete nightmare. This is particularly true at later stages when different molecular properties beyond potency must be optimized. Paradoxically, although activity cliffs are associated with structure/activity information, the corresponding knowledge might not be readily exploited when a given compound needs to be optimized. I think that activity cliffs really challenge the capabilities of computational approaches because the latter rarely provide good answers as far as cliffs is concerned. In broader terms, we should apply this concept in other disciplines such as catalysis. I don’t think we emphasize the cliffhanging culprits enough when we study the structure/activity relationships in ligand design.