Use of the supermolecule approach to derive molecular similarity descriptors for QSAR analysis

A relevant aspect in quantitative structure-activity (QSAR) and structure-selectivity (QSSR) relationships studies is the choice of the most appropriate molecular descriptors both with respect to the molecular series considered and the known or hypothetical mechanism of drug action. We have recently shown that ad hoc derived size and shape descriptors have been successful to derive QSAR and QSSR models for alpha(1)-adrenergic antagonists, 5-HT1A serotoninergic receptor ligands and M-1 muscarinic ligands, especially when dealing with non congeneric series of molecules. These descriptors describe the size-shape similarity with respect to a reference supermolecule which is obtained by superposition of the most active (selective) and structural different compounds, better if rigids. Molecular similarity indices based on molecular electrostatic potential (MEP) have found, as well, widespread use in the QSAR area. In the present study we extend the use of the supermolecule as a reference structure also in the context of MEP similarity. We have defined an ad hoc MEP similarity index with respect to the supermolecule using the same formalism of Hodgkin and Richards. The MEP of the supermolecule is computed as the average MEP of the compounds defining the supermolecule. A FORTRAN code is implemented to optimize the superposition of the ligands on the reference supermolecule in order to maximize the values of the ad hoc similarity descriptors defined in this study. The performance of the different matching criteria and the different ad hoc molecular similarity indices derived with the supermolecule approach are tested in QSAR modeling of the binding affinity and efficacy of a wide ranging series of M-1 muscarinic ligands previously studied.