Complementing ultrafast shape recognition with an optical isomerism descriptor

Abstract

We introduce the mixed product of three vectors spanning four molecular locations as a descriptor of optical isomerism. This descriptor is very efficient as it does not require molecular superposition, and is very robust in discriminating between a given isomer and its mirror image. In particular, conformational isomers that are mirror images of each other, as well as optical isomers have opposite sign of the descriptor value. For efficient database searches, the optical isomerism descriptor can be used to complement an available ultrafast shape recognition (USR) method based solely on distances, which is not able to distinguish enantiomers. By an extensive comparison of the USR-based similarity score with an approach based on Gaussian molecular volume overlap, the accuracy and completeness of the former are discussed.