Docking engines

The virtual_screening workflow supports two docking engines, selected with --docking_engine. Both use the same pose-sampling algorithm and the same docking box, so their results are directly comparable.

Engine	Scoring	Units (lower/higher better)	Relative speed
vina (default)	Empirical function	kcal/mol, lower is better	Fast
gnina	Empirical function + CNN rescoring	CNNaffinity (pK), higher is better	Slower, GPU recommended

Use vina for a fast baseline with no extra setup. Use gnina when you have a GPU (or can tolerate slower CPU runs) and want a learned pose/affinity score on top.

See the virtual screening workflow for the CLI flags, and the pages above for the scientific background on each engine’s scoring function.