Docking against multiple conformations of a single protein, i.e. ensemble docking, is a well established approach to move beyond the rigid body approximation and approximately include protein flexibility in docking. To facilitate more widespread application of ensemble docking, we introduce a method that automatically selects a small collection of protein conformations for docking from a larger ensemble of protein structures and generates a per-protein structure offset to combine per-ligand docking results into a single ranked list of compounds. This offset can be thought of as a rough estimate of the relative protein reorganization energy for each protein in the ensemble. Training per target employs a collection of known active ligands, a collection of presumed decoy compounds, and set of prepared protein structures. Excellent early enrichment performance is observed when using automated ensemble selection and offset estimation in ensemble docking against a diverse range of protein targets. Enrichment performance is compared to collections of randomly generated ensembles, ensembles of conformations generated by molecular dynamics simulations and Induced Fit Docking, ensembles cherry-picked from clusters formed by binding site 3D similarity, and single protein conformation docking. Guidance is provided to apply the method to other situations where multiple protein structures are common such as multiple crystal water molecules and uncertain protein tautomeric states.

Improving reliability of ensemble docking through automated ensemble selection and protein reorganization energy estimation

256th National Meeting (2018)

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