Evaluating molecular similarity using reduced representations of the electron density.

A model system of four benzodiazepine-like ligands for the central benzodiazepine receptors (CBRs) and peripheral benzodiazepine receptors (PBRs)is examined using a genetic algorithm procedure (GAGS) designed for evaluating molecular similarity. The method is based on the alignment of reduced representations generated from the critical points of the electron density computed at medium crystallographic resolution. The results are further characterized by a comparison with alignments produced by MIMIC, a field-based superimposition method that matches both steric and electrostatic molecular fields. The alignments produced by the two methods are generally seen to be consistent. The relationships of the compounds' binding affinities for both CBRs and PBRs to the alignments determined by GAGS yield a set of structural features required for significant binding to benzodiazepine receptors. Benefits of using reduced representations for evaluating molecular similarities and for constructing pharmacophore models are discussed.