To combat a coronavirus disease 2019 (COVID-19) caused by a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) safe, efficacious and specific antivirals are urgently needed. One promising drug target is a papain-like protease (PLpro), a viral protein essential for SARS-CoV-2 proliferation cycle. Several congeneric naphthalene-based small-molecule compounds were recently shown to inhibit both, PLpro activity in vitro, and SARS-CoV-2 replication in cells. In X-ray crystal structures these inhibitors occupy the same binding pocket resulting in disruption of peptide recognition. These data provide a starting point for structural-based drug design of PLpro inhibitors. <i>In silico</i> identification of potent inhibitors commonly requires prediction of protein-ligand binding free energy (BFE). Alchemical free energy methods such as thermodynamic integration (TI) allow for the accurate BFE estimation but are 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 TI calculations. First, we have selected approximately 10K commercially available small molecules with the same scaffold as known PLpro inhibitors. Selected ligands were then screened by an iterative procedure which combines relative BFE calculations by GPU-accelerated TI (using AMBER 18 package) and a machine learning model trained by using an active learning approach. At each iteration, 1) the ML model selects a set of ligands with the lowest predicted BFE; 2) TI calculations were performed for these ligands; 3) the calculated BFE were used to update the ML model. Using this procedure, we have identified a set of molecules with the lowest BFE which have been further tested experimentally. The details of the ligand set preparation, TI protocol, ML model as well as comparison of the calculations with experimental results will be reported.

Rational design of inhibitors for the SARS-CoV-2 papain-like protease using thermodynamic integration molecular dynamics guided by machine learning

ACS Fall 2021

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