Local intersection volume: a new 3D descriptor applied to develop a 3D-QSAR pharmacophore model for benzodiazepine receptor ligands

Abstract

In this work, we have developed a new descriptor, named local intersection volume (LIV), in order to compose a 3D-QSAR pharmacophore model for benzodiazepine receptor ligands. The LIV can be classified as a 3D local shape descriptor in contraposition to the global shape descriptors. We have selected from the literature 49 non-benzodiazepine compounds as a training data set and the model was obtained and evaluated by genetic algorithms (GA) and partial least-squares (PLS) methods using LIVs as descriptors. The LIV 3D-QSAR model has a good predictive capacity according the cross-validation test by ‘leave-one-out’ procedure ( Q 2=0.72). The developed model was compared to a comprehensive and extensive SAR pharmacophore model, recently proposed by Cook and co-workers, for benzodiazepine receptor ligands [J. Med. Chem. 43 (2000) 71]. It showed a relevant correlation with the pharmacophore groups pointed out in that work. Our LIV 3D-QSAR model was also able to predict affinity values for a series of nine compounds (test data set) that was not included into the training data set.