| name | autodock-vina-docking |
| description | Run reproducible local protein-ligand docking with AutoDock Vina and Meeko. Use for pose generation, redocking, or small virtual screens when a receptor and a justified binding pocket are available; do not use a Vina score as experimental affinity. |
| license | MIT |
AutoDock Vina Docking
Gate and preflight
Follow $cx-compute-environment and ask once before installing pinned Vina,
Meeko, or preparation dependencies and before compute. Require
$cx-molecular-input-preparation outputs and an explicit pocket definition.
Workflow
- Record Vina/Meeko versions, receptor state, ligand microstates, box center and
size in Å, exhaustiveness, modes, energy range, CPU count, and seed.
- Smoke-test one ligand. Confirm the search box encloses the intended pocket and
each ligand can rotate without clipping.
- Run a positive-control redock before screening. Keep raw PDBQT, log, config,
and converted SDF poses under
artifacts/<run-id>/vina/.
- For a screen, process each ligand state independently and preserve failures;
never silently keep only successful molecules.
- Rank poses and compounds separately. Report pose diversity and interactions,
then apply
$cx-docking-validation.
Interpretation
- Vina energy is a model score, not a measured binding free energy or IC50/Kd.
- Pose success requires a prespecified criterion, commonly top-k symmetry-aware
heavy-atom RMSD to a reference pose when that reference is valid.
- Use cold-target/ligand controls and decoys for comparative screening claims.
- Record apo/holo and assembly/template provenance; do not leak a held-out bound
pose into pocket construction or ligand initialization.