Molecular docking with AutoDock Vina: a step-by-step tutorial

AutoDock Vina is one of the most widely used free programs for molecular docking: predicting how a small molecule binds to a protein. This tutorial walks through a complete Vina run, from choosing a structure to reading the output, and shows how to check that your protocol works before you trust its predictions.

What you need

• AutoDock Vina (version 1.2 or later), free and open source.
• Open Babel or Meeko to prepare ligands, and a viewer such as PyMOL or UCSF ChimeraX.
• A protein structure from the RCSB PDB, ideally one with a bound ligand in the site you care about.

In this example we use HIV-1 protease bound to the inhibitor indinavir, PDB entry 1HSG, a classic teaching system because its binding mode is well known.

Step 1: choose and inspect the structure

Before docking, look at the structure:

• Resolution: better than about 2.5 Å is preferable for docking.
• Completeness: check for missing residues or side chains near the binding site.
• Ligands and waters: note the co-crystallised ligand, which defines the binding site and lets you validate the protocol.
• Biological assembly: HIV-1 protease is a homodimer and the binding site lies between the two chains, so keep both.

Step 2: prepare the receptor

1. Remove the co-crystallised ligand (save it separately; you will need it for validation) and, in most cases, water molecules.
2. Add hydrogens appropriate to physiological pH and check the protonation of key residues. In HIV-1 protease, the catalytic aspartates (Asp25 and Asp25′) deserve particular attention.
3. Convert the receptor to PDBQT format, which adds atom types and partial charges, using a preparation tool such as Meeko or AutoDockTools.

Step 3: prepare the ligand

1. Start from a reliable structure, such as an SDF from PubChem or the extracted crystal ligand.
2. Add hydrogens, generate a sensible 3D conformation and assign rotatable bonds.
3. Convert to PDBQT.

obabel ligand.sdf -O ligand.pdbqt -h --gen3d

Tools such as Meeko handle protonation and torsions more carefully than a one-line conversion, which matters for drug-like molecules.

Step 4: define the search box

Vina searches within a box. Centre it on the binding site, using the coordinates of the co-crystallised ligand, and make it large enough to contain the ligand with room to rotate, typically 20–25 Å per side for a drug-sized molecule.

receptor = receptor.pdbqt
ligand = ligand.pdbqt
center_x = 13.0
center_y = 22.5
center_z = 5.5
size_x = 22
size_y = 22
size_z = 22
exhaustiveness = 16
num_modes = 9

The centre values above are illustrative; compute them from the ligand coordinates in your own file.

Step 5: run Vina

vina --config config.txt --out docked.pdbqt

Vina reports a table of binding modes ranked by its scoring function, in kcal/mol, together with RMSD values between modes.

Step 6: validate by redocking

The single most important check is redocking: dock the crystal ligand back into its own receptor and measure the root-mean-square deviation (RMSD) between the top-scoring pose and the crystal pose. An RMSD below about 2 Å is the usual criterion that the protocol can reproduce a known binding mode.

If redocking fails, revisit the receptor preparation, protonation states, box size and exhaustiveness before docking anything new.

Step 7: inspect the interactions

Open the receptor and the top poses in PyMOL or ChimeraX and examine:

• hydrogen bonds to key residues (for HIV-1 protease, the catalytic aspartates and the flap region),
• hydrophobic contacts and buried surface,
• clashes or unsatisfied polar groups.

A pose with a good score but implausible chemistry should not be trusted.

What docking scores can and cannot tell you

• Vina scores are approximate and are best used to rank poses of one ligand and to prioritise compounds, not as accurate binding affinities.
• Comparing scores between very different molecules is unreliable; validate with actives versus decoys enrichment if you plan a virtual screen.
• Docking usually treats the receptor as rigid. Flexible regions, induced fit and water-mediated interactions are simplified.
• Promising poses can be tested further with molecular dynamics, and ultimately predictions need experimental confirmation.

For a deeper discussion of validation, see our guide to validating a docking study.

Go further

To learn docking properly, including target preparation, virtual screening and validation, see Molecular Docking & Structure-Based Drug Discovery, or try the short online Molecular Docking with AutoDock Vina workshop. You can inspect protein–ligand interactions for free in Nayancode Structure Studio.

Written by the Vivofold team for research and education. Found an error? Tell us and we will correct it.