Studies in Ligand Unbinding Transition State Plasticity for Kinetics-Oriented Drug Design

The dominant objective in drug design historically has been to improve drug efficacy through increasing affinity of drug binding under equilibrium conditions. However, a complete understanding of drug efficacy in non-equilibrium living organisms requires knowledge of the kinetics of drug (un)binding. While kinetics-oriented drug design has gained popularity it is still hampered by a number of limitations, not limited to the availability of structural models for ligand (un)binding transition states. In general these kinds of simulations are very difficult to achieve due to the long natural timescales of these processes (seconds to minutes) compared to the short timescales at which MD is computed (femtoseconds). In this thesis we address these limitations through computational methods for simulating full, unbiased, unbinding trajectories of inhibitors of drug targets with clinical interest. This is accomplished primarily by applying an enhanced sampling technique, called weighted ensemble (WE), over classical molecular dynamics (MD) simulations. Our approach is drastically more efficient than brute-force simulation methods, requires no biasing forces or other force field modifications, and is shown to work for a variety of systems of interest. Using these methods we are able to model, at all-atom resolution, the structure of unstable transition states for inhibitors of clinical interest of the soluble epoxide hydrolase (sEH) enzyme. This enzyme is implicated in a number of therapeutics including treatment of diabetic neuropathic pain. Critically, we also investigate the role of transition state plasticity in lead optimization. Towards this we developed a model for predicting plasticity from experimental data and a strategy for verifying these predictions which was applied in the context of sEH lead optimization.