Absolute Binding Free Energies with OneOPES

The calculation of\nabsolute binding free energies (ABFEs) for protein–ligand\nsystems has long been a challenge. Recently, refined force fields\nand algorithms have improved the quality of the ABFE calculations.\nHowever, achieving the level of accuracy required to inform drug discovery\nefforts remains difficult. Here, we present a transferable enhanced\nsampling strategy to accurately calculate absolute binding free energies\nusing OneOPES with simple geometric collective variables. We tested\nthe strategy on two protein targets, BRD4 and Hsp90, complexed with\na total of 17 chemically diverse ligands, including both molecular\nfragments and drug-like molecules. Our results show that OneOPES accurately\npredicts protein–ligand binding affinities with a mean unsigned\nerror within 1 kcal mol<sup>–1</sup> of experimentally determined\nfree energies, without the need to tailor the collective variables\nto each system. Furthermore, our strategy effectively samples different\nligand binding modes and consistently matches the experimentally determined\nstructures regardless of the initial protein–ligand configuration.\nOur results suggest that the proposed OneOPES strategy can be used\nto inform lead optimization campaigns in drug discovery and to study\nprotein–ligand binding and unbinding mechanisms.