Preparation and Implementation of a High Throughput Virtual Screening Protocol on a Shared Memory GPU Supercomputer

Abstract

A novel computational protocol for drug discovery is presented. Pre-screening binding sites of selected targets with high accuracy and precision using different molecular docking programs is combined with a virtual drug discovery platform (VDDP) to screen a large number of small molecules. Several relevant proteins were identified as targets for this protocol. The protein coordinates were downloaded for docking studies from the RCSB PDB for each target. FBPase was used as a case study for the complete implementation of the VDDP. Compounds with high affinity predicted scores were identified and molecular dynamics studies were performed on the 10 highest scoring protein/small molecule complexes focusing on the allosteric binding sites. Studies were carried out using NAMD. Dissociation constants (Ki’s) for the protein–ligand complex were calculated throughout the NAMD runs and respective conformational changes in the FBPase binding pocket were observed. The inhibitor binding pocket was evaluated based on pocket shape and volume. Predicted interfacial structural changes produced in response to small molecule binding at two allosteric sites was generated in NAMD. Novel hydrogen bonding and hydrophobic interaction networks of residues in the 3D structures were identified to connect the active sites to the allosteric binding sites using an NAMD protocol.