Glossary

The score AutoDock Vina returns for each pose. More negative is better — a lower (more negative) number means a more energetically favourable interaction between ligand and receptor.

These thresholds are rules of thumb, not laws. Affinity for drug-like molecules typically clusters around −7 to −9 kcal/mol; metallic cofactors and tight covalent inhibitors push lower, while large flexible ligands look worse than they really are because Vina rewards rigid hits.

How many independent search runs Vina performs. Higher numbers mean more thorough sampling of the conformational space — and slower runtime. The default of 8 is a reasonable trade-off for most undergraduate work; bump to 16 or 32 for publication-quality results, or for ligands with many rotatable bonds.

The maximum number of binding poses Vina returns. Even if you ask for nine, it only returns that many if it can find that many distinct poses within the energy range. Reduce to one if you only care about the best fit; raise to twenty if you're surveying the binding landscape.

The maximum kcal/mol gap between the best pose and the worst pose returned. With the default of 3.0, a job whose best pose is −8.5 will only return poses with affinities above about −5.5 kcal/mol. Tighten this (e.g., 1.0) if you only want very-similar poses; loosen it (e.g., 6.0) to see more alternative geometries.

The 3D rectangle in which Vina is allowed to place the ligand. MolDock computes the geometric centre of the protein by default — you almost always want to narrow this to a known binding pocket once you've got an initial result, so the search doesn't waste time on the surface. Box sizes between 18 and 25 Å per side cover most pockets.

AutoDock's extended PDB format. Each atom carries an additional Gasteiger partial charge plus an autodock atom type (e.g. C.ar for aromatic carbon). MolDock produces this automatically when you submit, but you can download the docked PDBQT and feed it back into other AutoDock-family tools.

The open-source docking engine MolDock wraps. Vina was published by Trott & Olson in 2010 and uses an iterated local search global optimiser with an empirical scoring function. It's fast (seconds to minutes per ligand) and good enough for triage / teaching, though dedicated physics-based methods (e.g. FEP) outperform it in absolute accuracy.