G-Protein Coupled Receptors Virtual Screening Using Genetic Algorithm Focused Chemical Space

Abstract

Exploiting the ever growing set of activity data for compounds against biological targets represents both a challenge and an opportunity for ligand-based virtual screening (LBVS). Because G-protein coupled receptors (GPCRs) represent a rich set of potential drug targets, we sought to develop an appropriate method to examine large sets of GPCR ligand information for both screening collection enhancement and hit expansion. To this end, we have implemented a modified version of BDACCS that removes highly correlated descriptors (rBDACCS). To test the hypothesis that a smaller, focused descriptor set would improve performance, we have extended rBDACCS by using a genetic algorithm (GA) to choose target-specific descriptors appropriate for selecting the set of 100 compounds most likely to be active from a decoy database. We have called this method GA-focused descriptor active space (GAFDAS). We compared the performce of rBDACCS and GAFDAS using a collection of activity data for 252 GPCR/ligand sets versus two decoy databases. While both methods appear effective in LBVS, overall GAFDAS performs better than rBDACCS in the early selection of compounds against both decoy databases.