Conformation Search Across Multiple-Level Potential-Energy Surfaces (CSAMP): A Strategy for Accurate Prediction of Protein–Ligand Binding Structures

Accurate protein binding structure determination presents a great challenge to both experiment and theory. Here, in this work, we propose a new DOX protocol which combines the ensemble molecular Docking as the coarse-level, structure Optimization with the semiempirical quantum mechanics methods as the medium level, and the eXtended ONIOM (<i>X</i>O) calculations as the fine level. The fundamental of the DOX protocol relies on the Conformation Search Across Multiple-level Potential-energy surfaces (CSAMP) strategy, where the conformation spaces of a funnel-like structure are searched from the coarse level with hundreds of candidates to the medium level with around 10 top candidates to the fine level with the final top 1 or 2 binding modes. An in-depth test for the protocol set up against 28 crystallographic data consisting of HMGR-statins, SDase-inhibitors, 3HNRase-inhibitors, and NA-inhibitors yielded a satisfactory result with ∼0.5 Å root-mean-square deviations (RMSDs) on geometries and ∼0.8 kcal/mol absolute error of relative binding energies on average. A further larger scale validation on the Astex test set (including 85 diverse structures) revealed an impressive performance with a RMSD < 2 Å success rate of 99%, suggesting DOX is a promising computational route toward accurate prediction of the protein–ligand binding structures.