ModBind, a Rapid Simulation-Based Predictor of Ligand Binding and Off-Rates.

Abstract

In rational drug discovery, both free energy of binding and the binding half-life (k_off) are important factors in determining the efficacy of drugs. Numerous computational methods have been developed to predict these important properties, many of which rely on molecular dynamics (MD) simulations. While binding free-energy methods (thermodynamic equilibrium predictions) have been well validated and have demonstrated the ability to drive daily synthesis decisions in a commercial drug discovery setting, the prediction of k_off (kinetics predictions) has had limited validation, and predictive methods have largely not been deployed in drug discovery settings. We developed ModBind, a novel method for MD simulation-based k_off predictions. ModBind demonstrated similar accuracy to current state-of-the-art free-energy prediction methods. Additionally, ModBind performs ∼100 times faster than most available MD simulation-based free-energy or k_off methods, allowing for widespread use by the molecular modeling community. While most free-energy methods rely on relative free-energy changes and are primarily useful for optimization of a congeneric series, our method requires no structural similarity between ligands, making ModBind an absolute predictor of k_off. ModBind is thus a tool that can be used in virtual screening of diverse ligands, making it distinct from relative free-energy methods. We also discuss conditions that enable approximate prediction of ligand efficacy using ModBind and the limitations of this approach.