Computational screening is a method to rapidly prioritize small molecules based on defined or predicted structural and biochemical attributes built from ligand and target information. Previously, we have developed a highly scalable virtual screening workflow of kinase activity predictors and parallel ensemble docking models to identify novel dual kinase bromodomain inhibitors. Bromodomains are epigenetic reader proteins that recognize and bind specific acetylation marks on histones, facilitating chromatin decondensation and promoting transcription initiation. Bromodomain-containing protein 4 (BRD4) has been linked to the expression of many oncogenes such as MYC and has become a promising target in cancer therapy. From our previous study we identified several novel oxopyrrolidinylphenyl-aryl-sulfonamide compounds that scored high in our ensemble docking with data fusion protocol. In this study we quantified the binding potential of these compounds biochemically and further improved our docking models by performing 50 nano-second (ns) all atom explicit water molecular dynamics simulations with those that displayed binding activity. The molecular dynamics systems were generated from the protein-ligand complex corresponding to the highest docking score for each active ligand. MD simulations worked well with protein backbone RMSD stabilizing within 2 Å after 10 ns and ligand RMSD from the starting docking conformation increasing to 4 Å after 20 ns. Crystal structures of six active compounds from this scaffold binding BRD4 were determined. Our molecular dynamics protocol was then repeated for each crystal structure and no significant deviations from the co-crystal pose were obtained within the simulation time. Each active ligands molecular dynamics simulation trajectory was analyzed starting from its highest-scoring docking pose and using the determined co-crystal structure as reference. Our results show that the MD simulations can reproduce the experimentally observed co-crystal structure protein ligand complex for all six compounds. Although the initial high-scoring docking poses do not exactly align with the co-crystal structure ligand binding configuration, MD simulations were able to generate binding models that correspond to experimental structures within 20-30 ns simulation time. We believe that our high-throughput computational structural biology approach is applicable to all realms of drug discovery and can be of great value to the scientific community.

Identification of a novel class of BRD4 inhibitors by high-throughput computational structural biology

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.