THE MOLECULAR TRANSFORM AS A SIMILARITY MEASURE

We propose to quantify molecular similarity through various forms of molecular transforms directly related to experimental measurements. Various metric distances between molecular transforms are introduced in measuring similarity which can be used in quantitative structure-activity relationships. For simpler classes of compounds like aliphatic alcohols good correlations are obtained between the abstract distance from a lead compound and various physical and pharmalogical properties. For substituted phenols the correlation is worse; however, the predictive power of the descriptors derived from the molecular transform is yet acceptable. For trypsin inhibitors, a class of compounds having very different molecular formulae, the net atomic charge is introduced as a parameter in the generalized form of the molecular transform. Though a poor regression equation is obtained for the differences in the inhibitory power, inactive compounds within a set can be reliably selected.