Protein kinases have become important targets for drug discovery. Over 30 small-molecule inhibitors have now been approved as therapeutic drugs. As the family of human protein kinases numbers over 500 and various mutants are responsible for causing diseases, many more drugs could be developed to target more protein kinases. Targeting protein kinases of microbes has also been explored. This project continues to refine and add tools to a virtual screening pipeline to speed up the identification of useful drug candidates for these proteins, with refining ensemble docking and adding a component on drug-binding kinetics as major goals. Ensemble docking provides a fast method for incorporating receptor flexibility in virtual screening. Unfortunately, researchers have not yet found an effective way to use the docking scores from multiple structures of an ensemble to select active compounds. Researchers have also been puzzled by the decrease rather than increase in performance as more structures are added. This proposed research carries out a larger-scale study to further validate a Bayesian model that has been shown to resolve these problems. Drug-binding kinetics is now well recognized as an important factor to be considered early in a drug-discovery process. To predict drug-binding kinetics and derive structure-kinetics relationships to guide drug design, this research employs three major methods with different strengths. Fast steered molecular dynamics is first used to screen a large set of compounds. The most promising drug candidates are then re-confirmed by the slower but more rigorous Markov state model and the mile-stoning method before suggesting the most hopeful compounds to the experimentalists for testing. Using several methods with different approximations to check consistency will help to draw more reliable conclusions. This research will also extend the PyMile package to perform exact mile-stoning simulations to add another level of confirmation.
Specific Aim 1 of this application performs a larger-scale study to benchmark the promising Bayesian model for ensemble docking.
Specific Aim 2 performs a larger-scale study to further benchmark steered molecular dynamics simulations? capability to rapidly select drug candidates with long residence times from chemical libraries. It continues to evaluate and improve the Markov state model and the mile-stoning method for calculating the absolute dissociation rates of ATP-competitive inhibitors from protein kinases. To help develop structure-kinetics relationships, this aim also examines the generality of a two-step mechanism of dissociation of these inhibitors and investigates how slow- and fast-dissociating ligands use this two-step mechanism differently. The proposed studies will provide a more robust model of ensemble docking for virtual screening, and a faster practical approach for finding drug candidates with therapeutically useful kinetics from compound libraries. Both will improve scientists? ability to zoom into good drug candidates faster.
Over 30 small-molecule inhibitors of protein kinases have been approved as therapeutic drugs, mostly for treating cancers. As the family of protein kinases contains over 500 members and various mutants are responsible for causing diseases, many more drugs could be developed to target more protein kinases. The proposed research improves tools of computer-aided drug design to speed up the discovery of drugs targeting this family of proteins.