Are protein–ligand docking programs good enough to predict experimental poses of noncovalent ligands bound to the SARS-CoV-2 main protease?

Abstract

Hundreds of virtual screening (VS) studies have targeted the severe acute respiratory syndrome-coronavirus 2 (SARS-CoV-2) main protease (M-pro) to identify small molecules that inhibit its proteolytic action. Most studies use AutoDock Vina or Glide methodologies [high-throughput VS (HTVS), standard precision (SP), or extra precision (XP)], independently or in a VS workflow. Moreover, the Protein Data Bank (PDB) includes multiple complexes between M-pro and various noncovalent ligands, providing an excellent benchmark for assessing the predictive capabilities of docking programs. Here, we analyze the ability of the three Glide methodologies and AutoDock Vina by using various target structures/preparations to predict the experimental poses of these complexes. Our aims are to optimize target setup and docking methodologies, minimize false positives, and maximize the identification of various chemotypes in a SARS-CoV-2 M-pro noncovalent inhibitor VS campaign.