Automatization of the Nwat-MMGBSA method to rescore docking results in medium-throughput virtual screening applications

Nwat-MMGBSA is a variant of MM-PB/GBSA based on the inclusion of a number of explicit water molecules that are the closest to the ligand in each frame of a molecular dynamics trajectory.[1] This method can increase the correlation between predicted and experimental binding energies in both ligand-receptor and protein-protein complexes,[2] compared to standard MM-GBSA. The protocol for molecular dynamic (MD) simulations, preparatory to subsequent Nwat-MMGBSA calculations, has now been optimized to maximize efficacy and efficiency, thus making the calculations practical in low-to-medium throughput virtual screenings. Three systems, penicillopepsin, HIV1-protease and BCL-XL, have been used as test cases. Calculations have been performed in triplicates on both classic HPC environments as well as on workstations equipped by GPU cards, evidencing no statistical differences in the results, but a dramatic decrease of the “cost per nanosecond” for the latter systems. With the optimized protocol, the whole calculation, from equilibration to production MD and subsequent Nwat-MMGBSA rescoring, averagely took about one hour per ligand using a single GPU. A set of scripts for automatic structure based virtual screening, from library setup to docking and rescoring, has also been designed and tested within a retrospective virtual screening for inhibitors of the Rac1-Tiam1 protein-protein interaction. The screening library has been built using compounds experimentally tested, with a ratio between actives and real inactives of 1 to 10, and different protocols were used to process the library prior to docking (DB-A, DB-B and DB-C, Figure 1). The results, summarized in Figure 1, confirmed the benefit of including explicit water molecules MM-GBSA calculation and the validity of Nwat-MMGBSA to rescore virtual screening results.