What is molecular docking?

Molecular docking is a computational method that predicts the preferred position, orientation and shape — the pose — of a molecule (the ligand) when it binds to a protein (the receptor). It also gives each pose a score that estimates how strongly the two interact.

How docking works

Every docking program combines two components:

• A search algorithm explores possible poses: moving and rotating the ligand within the binding site and twisting its rotatable bonds. Classical tools such as AutoDock Vina use stochastic optimisation; newer deep-learning tools such as DiffDock generate poses with a trained model.
• A scoring function ranks the poses. Physics-based, empirical and knowledge-based functions estimate contributions such as hydrogen bonds, hydrophobic contacts, steric clashes and the cost of freezing flexible bonds. See docking scores.

Inputs and outputs

1. A protein structure, from the Protein Data Bank or from structure prediction, prepared by adding hydrogens and removing unwanted water and ions.
2. A binding site, usually defined as a 3D box around a known pocket. “Blind” docking searches the whole protein instead. See binding sites.
3. Ligands, typically supplied as SMILES and converted to 3D with sensible protonation states.

The output is a ranked list of poses per ligand, each with a score, usually viewed in a 3D molecular viewer.

How reliable is docking?

Docking is good at pose prediction when the protein’s binding site is well defined: re-docking a ligand into its own crystal structure often reproduces the experimental pose within 2 Å RMSD. It is much weaker at predicting affinity. Scores correlate only loosely with measured binding, so docking is best used to enrich a list of candidates rather than to rank close analogues precisely.

Common pitfalls include rigid protein models that cannot adapt to the ligand, missing water molecules that mediate binding, wrong protonation or tautomer states, and very flexible ligands such as large macrocycles, whose ring conformations must be sampled carefully.

Good practice

• Validate the setup first by re-docking a known ligand.
• Inspect the top poses visually: do they make chemically sensible contacts?
• Compare scores within a series, not across unrelated targets.
• Confirm promising compounds experimentally — docking generates hypotheses, not answers.