Discovering Dengue Drugs-Together: Grid Scale Virtual Screening with Mean Field Free Energy of Binding Rescoring

Abstract

The virtual screening of chemical databases against drug discovery targets with docking programs can enrich a database for bioactive compounds. However, current virtual screening methods generate many false positives leading to extensive and expensive testing of ultimately inactive compounds. In addition, the performance of virtual screening methods is dependent on the target system. This work examines if coupling traditional docking based virtual screening methods with perturbation based mean field free energy of binding (MF-FEB) calculations to rescore docking generated poses will improve enrichment over traditional virtual screening methods. MF-FEB calculations are computationally demanding requiring the distributed computing resources of IBM’s World Community Grid. The work details three retrospective studies of MF-FEB using a 30 compound test set for binders and non-binders of the L99A T4 lysozyme, the DUD estrogen agonist test set and the DUD trypsin test set. In addition, this work describes the active prospective drug discovery effort on World Community Grid Discovering Dengue Drugs-Together that utilizes MF-FEB rescoring. The testing MF-FEB rescoring showed that while MF-FEB calculations can improve enrichment over traditional virtual screening methods it still has many of the same limitations as traditional virtual screening methods.