AutoDock Vina 1.2.0: new docking methods, expanded force field, and Python bindings

AutoDock Vina 1.2.0 extends the original Vina engine (Trott & Olson 2010) with macrocycle conformational sampling, hydrated docking protocol, AutoDock4.2 scoring function support, simultaneous multi-ligand docking, batch mode, and Python bindings. The benchmarks in this paper use DUD-E (102 targets), not CASF-2016.

The paper establishes Vina 1.2.0 as the canonical software version for subsequent docking pipelines that use these features. It does NOT define a kcal/mol score threshold for hit identification.

Key finding

New features improve specific use cases (macrocycles, bridging waters) but do not change the Vina scoring function for rigid standard docking. For h01 Phase 4b, which uses rigid receptor + standard Vina scoring, the relevant performance baseline remains the original Trott & Olson 2010 benchmark.

Numbers that matter

MetricValueContext
DUD-E top-1 success rate (Vina, 2 Å RMSD)68%102-target DUD-E dataset
DUD-E top-2 success rate (Vina, 2 Å RMSD)78%
DUD-E top-3 success rate (Vina, 2 Å RMSD)84%
AUC (DUD-E, Vina scoring function)0.72 ± 0.12Virtual screening enrichment
Score threshold for hit classificationNot defined in paperNo kcal/mol cutoff given
Exhaustiveness recommendation64 for macrocyclesHigher than default 8

What this paper does NOT provide

  • No CASF-2016 benchmark (uses DUD-E only).
  • No kcal/mol score threshold for hit identification.
  • No recommendation for a “passing” Vina score. The −5.0 kcal/mol gate in h01 Phase 4b is a pipeline-specific positive-control gate (diflunisal benchmark), not derived from this paper.

Limitations

  • DUD-E has known dataset bias (artificial decoys inflate enrichment metrics).
  • Performance is target-dependent: BEDROC < 0.1 for 25/102 targets.
  • Scoring function unchanged from Vina 1.1.2 for standard (non-macrocycle, non-hydrated) docking.

Connections