How to run molecular docking, step by step
A practical workflow for protein–ligand docking, from cleaning the structure to judging the results.
What is molecular docking?
Molecular docking is a computational method that predicts how a small molecule (ligand) fits into the binding site of a protein and estimates how favourable that fit is. It is widely used in early drug discovery to prioritise compounds before experiments.
The workflow
- Choose and clean the protein structure. Download a high-resolution structure from the Protein Data Bank. Remove unneeded water, ions and duplicate chains, and keep the co-crystallised ligand for validation.
- Prepare the receptor. Add hydrogens, assign protonation states and partial charges, and repair missing side chains or loops where needed.
- Prepare the ligands. Generate 3D structures, set protonation and tautomer states at the target pH, minimise the energy and convert to the docking format.
- Define the binding site. Place a grid box around the known active site or the co-crystallised ligand. It should be large enough for the ligand to move, but not so large that search quality drops.
- Validate the protocol. Re-dock the native ligand and compare it to the crystal pose. An RMSD below about 2 Å suggests the settings are reasonable.
- Run the docking. Dock your compound set with sufficient search exhaustiveness and save several poses per ligand.
- Analyse the results. Rank by score, then inspect hydrogen bonds, hydrophobic contacts and pi interactions, and compare with a reference inhibitor.
Common mistakes
- Skipping validation, so nobody knows whether the settings can reproduce a real pose.
- Wrong protonation states for the ligand or key residues.
- A grid box that is not centred on the true binding site.
- Ranking by score alone and ignoring chemistry and interactions.
What to do after docking
Promising hits are usually taken forward to molecular dynamics simulation to test whether the complex stays stable, followed by MM/PBSA free-energy estimates and laboratory testing.
Frequently asked questions
Which software is used for molecular docking?
Common choices include AutoDock Vina, AutoDock4, Glide and GOLD. The best choice depends on the target, the ligand type and the available licences.
What is a good docking score?
There is no universal cut-off. Scores are relative, so compare compounds with a known reference ligand on the same receptor and grid, then inspect the poses.
How do I validate a docking protocol?
Re-dock the co-crystallised ligand and measure the RMSD between the docked and crystal pose. Below about 2 Å is generally considered acceptable.
Does a good docking score prove a drug works?
No. Docking predicts a plausible pose and relative ranking. Follow it with molecular dynamics, binding free-energy estimates and experiments.
Related guides
Build topologies, solvate and neutralise, minimise, equilibrate in NVT then NPT, run production, then correct periodic boundaries before analysis.
Virtual screening guideVirtual Screening Workflow: From Library to Hit ListPrepare the target and library, filter by properties, validate the docking setup, dock, then rank, cluster and inspect before choosing compounds to test.
Related service
Written by Sohel Akhtar Ansari, ReactorMD founder (ORCID). Computational predictions are not experimental confirmation.