Standardization of virtual-screening and post-processing protocols relevant to in-silico drug discovery

Structure-based drug discovery has emerged as a powerful tool in computational drug discovery and has gained rapid acceleration due to the development of better algorithms for high-end computation in an affordable time. Molecular docking and virtual screening methods are routinely used for the purpose but computing the ligand binding energies with inbuilt scoring functions accurately is still a limitation. Although, MMPBSA and MMGBSA are routinely employed tools for achieving accurate binding free energies, they are applied on well-equilibrated explicitly solvated systems and are computationally expensive and time-consuming. This study compares different post-processing protocols performed on an in-silico screened benchmarked P. falciparum Dihydrofolate reductase (PfDHFR) dataset with AutoDockVina. The docked and implicitly solvated complexes were subjected to (1) rescoring, (2) energy minimization and (3) Binding Estimation After Refinement (BEAR) algorithm. Subsequently, binding free energies were computed using three different tools-MMPBSA, MMGBSA and "g_mmpbsa". Surprisingly, rescoring alone displays lower accuracy than the inherent scoring function of the AutoDockVina. However, encouraging results were seen after post-processing with the other two protocols. The results suggest that MMPBSA applied on energy minimized conformations is able to achieve 42-fold reductions in computational time as opposed to the BEAR algorithm with comparable accuracy.