Ligand binding affinity optimization is one of the major challenges in computational drug design. Alchemical binding free energy (BFE) calculations, which utilize molecular dynamics (MD) simulations in explicit solvent, can provide accurate predictions of protein-ligand binding affinity. However, computing BFE for a large number of ligands with this method is computationally expensive and time-consuming. In this work, we have developed an efficient automated workflow for identifying compounds with the lowest BFE among thousands of congeneric ligands which requires only hundreds of alchemical calculations. The workflow includes alchemical free energy and active learning combined in the iterative procedure. At each iteration, 1) the machine learning (ML) model selects a set of ligands with the lowest predicted BFE; 2) alchemical BFE calculations are performed for these ligands; 3) the calculated BFEs are used to update the ML model. The workflow was applied to identify small-molecule inhibitors of SARS-CoV-2 papain-like protease (PLpro), a promising molecular target for the design of COVID-19 therapeutics. X-ray crystal structure of PLpro bound to a small-molecule inhibitor was used as the starting point. We virtually screened a library of 1.3 billion commercially available compounds, selected a focused library of approximately eight thousand compounds with the same and similar scaffolds as the inhibitor in the PLpro crystal structure, and finally identified sixteen potent binders with more than a hundred-fold improvement in predicted binding affinity. The details of the ligand set preparation, alchemical BFE calculation protocol, and ML model will be reported.

Combined alchemical free energy calculations and active learning for optimization of the SARS-CoV-2 papain-like protease small-molecule inhibitor

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

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