As the interest in drug-design for kinetically optimized inhibitors grows there will be increased interest by medicinal chemists for "crystal structures" of unbinding transition states. Experimental techniques in structural biology can not produce such 3-dimensional structures due to their transient existence. Thus, computational approaches (and molecular dynamics particularly) are poised to supply this need. However, the main limitation of MD simulations has been that they are inefficient at simulating rare events with timescales beyond the micro- and millisecond ranges. Recent developments in enhanced sampling methods, like metadynamics and weighted ensemble, have effectively shown that they can ameliorate this problem for many systems, including ligand unbinding. These achievements unlocks the potential to perform direct structural analyses of transition states, as well as providing feedstock for data-driven drug design pipelines oriented specifically for optimizing kinetics. The exciting technical achievement of these simulations still beg the question of when, how, or if at all this kind of data will be useful in drug-design. The principle questions will be: 1) how sensitive are ligand-unbinding transition states are to chemical perturbations and 2) what are the important features of ligand-unbinding transition states and are they even controllable through drug design. Here we use the weighted ensemble technique called WExplore to generate unbinding trajectories and transition state structures for a number of lead inhibitors for the protein, soluble epoxide hydrolase (sEH). We start with typical linear structure activity/kinetics relationships (SAR/SKR) models and Ki vs. kd plots used by medicinal chemists to predict the effect of chemical modifications. Using a library of inhibitors with experimentally known affinities and kinetics we then test predictions about the structure of their transition states. I.e. which perturbations to the chemical structure produce nonlinear effects to the structure of the transition state and unbinding pathway. We aim then to predict chemical space domains (molecular libraries) that maintain linearity within SKR models. These hardened models along with 3D structures of the transition states (and broader unbinding landscape) will provide valuable decision support analytics for medicinal chemists.

Predicting plasticity of ligand-unbinding transition states for a library of minutes-scale kd inhibitors

2019 Fall National Meeting

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