Similarity-based SIBAR descriptors for classification of chemically diverse hERG blockers

Abstract

There is an increasing interest in computational models for the classification and prediction of the human ether-a-go-go-related-gene (hERG) potassium channel affinity in the early phase of drug discovery and development. In this study, similarity-based SIBAR descriptors were applied in order to develop and validate in silico binary QSAR and counter-propagation neural network models for the classification of hERG activity. The SIBAR descriptors were calculated based on four reference datasets using four sets of 2D- and 3D-descriptors including 3D-grid-based VolSurf, 3D 'inductive' QSAR, Van der Waals surface area (P_VSA) and a set of 11 hERG relevant 2D descriptors devised from feature selection methods. The results indicate that the reference data set tailored to the specific problem, together with a set of hERG relevant descriptors, provides highly predictive models.