The impact of library size and scale of testing on virtual screening

Virtual ligand libraries for ligand discovery have recently increased 10,000-fold. Whether this has improved hit rates and potencies has not been directly tested. Meanwhile, typically only dozens of docking hits are assayed, clouding hit-rate interpretation. Here we docked a 1.7 billion-molecule virtual library against β-lactamase, testing 1,521 new molecules and comparing the results to a 99 million-molecule screen where 44 molecules were tested. In a larger screen, hit rates improved twofold, more scaffolds were discovered and potency improved. Fifty-fold more inhibitors were found, supporting the idea that the large libraries harbor many more ligands than are being tested. In sampling smaller sets from the 1,521, hit rates only converged when several hundred molecules were tested. Hit rates and affinities improved steadily with docking score. It may be that as the scale of docking libraries and their testing grows, both ligands and our ability to rank them will improve.