The Use of Procrustes Analysis to Compare Different Property Sets for the Characterization of a Diverse Set of Compounds

Abstract
Procrustes analysis has been used to compare the similarity of 14 different sets of property descriptors for characterizing a diverse set of pharmaceutical compounds. The analysis generated 3 distinct clusters from the property sets, which could be generally described as a) mostly 2D physicochemical properties, b) physicochemical properties based on overlaid 3D molecular structures, c) connectivity indices. Compound clustering and selection methods based on similarity measures calculated from a property set in one cluster would therefore give quite different results to those from a set in a different cluster.