Benchmark · Molecular docking · Draft
AutoDock Vina versus Vinardo on Astex Diverse.
With preparation, search space, seed, and exhaustiveness held constant, Vinardo generated more sub-ångström poses but did not improve top-ranked pose success.
Controlled comparison
Both scoring functions were evaluated across the same 85 Astex Diverse complexes using an identical ligand-extent-plus-4-ångström search box, seed 20260720, exhaustiveness 32, and ten output modes.
Primary result
Vina achieved 62.4% top-1 success at 2 Å versus 60.0% for Vinardo; Vinardo improved sub-1 Å success from 36.5% to 45.9% and median RMSD from 1.27 Å to 1.17 Å.
Interpretation
Both methods sampled a correct pose in more than 84% of cases. The remaining accuracy reserve is primarily a pose-ranking problem rather than a pose-generation problem.