Morphological similarity: A 3D molecular similarity method correlated withprotein-ligand recognition

Recognition of small molecules by proteins depends on three-dimensional mol ecular surface complementarity. However, the dominant techniques for analyz ing the similarity of small molecules are based on two-dimensional chemical structure, with such techniques often outperforming three-dimensional tech niques in side-by-side comparisons of correlation to biological activity. T his paper introduces a new molecular similarity method, termed morphologica l similarity (MS), that addresses the apparent paradox. Two sets of molecul e pairs are identified from a set of ligands whose protein-bound states are known crystallographically. Pairs that bind the same protein sites form th e first set, and pairs that bind different sites form the second. MS is sho wn to separate the two sets significantly better than a benchmark 2D simila rity technique. Further, MS agrees with crystallographic observation of bou nd ligand states, independent of information about bound states. MS is effi cient to compute and can be practically applied to large libraries of compo unds.