A recent paper has described a new computational method for chemical scaffold hopping by fragment replacement implemented and currently used by AstraZeneca. Vainio et al.’s method is simply termed Scaffold Hopping (SH) and it follows the CAVEAT approach of finding similar fragments to each of the non-overlapping fragments derived from the query molecule. The leading author had previously introduced ShaEP, a notable 3D similarity search method.
SH is powered by SSD, a variant of USR that is designed to identify scaffold hops that have at least one fragment attachment vector in common, and thus is shown to be able of searching a database with a billion fragment conformers in a couple of minutes using a single processing core. SH is found to be more effective at retrieving known bioisosteres than SSD or USR alone, which shows the improvement introduced by SH in those scaffold hops requiring at least one common substituent. However, USR has the advantage of being able to generate more distant scaffold hops, which is not possible by construction in SH or SSD because of the need to preserve at least one substituent. Prospective examples of the identification of distant scaffold hops using USR can be found here, here and here.
It is nice to see the methods one develops being used by pharmaceutical companies. It would be interesting to see how much recent pharmacophoric extensions of USR, such as the excellent work by Adrian Schreyer and Tom Blundell on USRCAT, can improve over the published results once adapted to the SH workflow.