Large-scale Validation of Drug Discovery Pipeline Protocols

Project Description

Drug discovery is a challenging and expensive endeavor where computational approaches can add considerable value. Our project will use ORNL’s compute resources to do research into several aspects of the computational aided drug discovery pipeline. As a first step of a drug discovery pipeline, millions of compounds are screened (or docked) against a target for low-energy “hits” that are then optimized into leads. Docking can comprise multiple programs with multiple parameters, and a thorough search of these different parameters to find their efficacy on the drug candidates has currently not been undertaken. Our research will use Bayesian optimization on analyze multiple docking simulations completed with various applications and parameters. This analysis could would reveal the optimal parameters from a drug discovery perspective. Another facet of the drug discovery pipeline involves prediction of the off target effects of a promising drug candidate. The action of a drug is directly related to its engagement with a target while undesirable side effects generally come from unintended binding to off-targets. To address the above concerns, we are testing a much more rigorous approach to predict protein-ligand binding affinity using free energy perturbation (FEP) with molecular dynamics simulations in explicit water molecules including full sampling of all relevant degrees of freeedom.However, due to the high computational resource requirements, limited work has been done to assess the predictive capabilities of FEP and even less work has been done exploring the numerous parameters and protocol that can be controlled. We would use similar techniques for docking protocol optimization to assess the capabilities of FEP as a lead optimization approach.