Cross-docking, the problem of docking a ligand into its own protein where the structure was crystallized in the presence of a different ligand, is highly relevant to the effective use of virtual screening in drug design, but remains a considerable challenge. Using induced charges, we have investigated the effect of including ligand and enzyme polarization in the protein-ligand cross-docking problem for 15 sets of complexes with at least 7 structures containing different drug-like ligands, resulting in 5238 cross-dockings; proteins were included if there was at least 99% sequence identity and at least 97% coverage. The initial docking was carried out using GLIDE. The ligand was then polarized using hybrid quantum mechanical / molecular mechanics methods using Jaguar and QSITE and the protein was polarized using the induced charge method, which required the electrostatic potential determined at the atoms within the enzyme. The polarized ligand was then re-docked using GLIDE into both the polarized enzyme and the unpolarized enzyme. The improvement in sampling with full polarization (ligand plus enzyme) is 5% for self-docking, 11% for cross-docking and 15% for mutual cross-docking, as judged by an RMSD of less than 2 Å. The effect of polarization on enrichment was assessed by docking 208 actives and 11404 carefully designed decoys to the adenosine A2A receptor. The top 3% of hits were re-docked using polarized docking, giving a marked increase in enrichment within the top 0.5-1.5% of hits in terms of both the number of actives. The effect of polarization on the ligand binding pathway was assessed using supervised MD simulations.

Addressing the cross-docking problem through polarization

2019 Spring National Meeting

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