Molecular Docking Tutorial — From PDB Files to Binding Results

What Is Molecular Docking?

Molecular docking is a computational method that predicts how a small molecule (ligand) binds to a target protein. It samples thousands of possible binding orientations and conformations, scoring each one using energy-based functions to identify the most favorable binding pose. Docking is a cornerstone of structure-based drug discovery, enabling virtual screening of millions of compounds before laboratory testing.

The fundamental principle is the lock-and-key model: the protein (lock) has a binding site (keyhole) where a ligand (key) can bind. Modern docking accounts for flexibility in both the protein and ligand, using scoring functions that estimate binding free energy. The lower the binding energy, the stronger the predicted interaction.

Molecular Docking Workflow: Step by Step

Step 1: Obtain Protein Structure (PDB File)

Download the protein 3D structure from the Protein Data Bank (PDB). Search by PDB ID, protein name, or gene. For proteins without experimental structures, use ESMFold or AlphaFold2 for predicted models. Ensure the structure covers the binding site of interest and has reasonable resolution (< 3.0 Å for X-ray, < 4.0 Å for cryo-EM).

Step 2: Prepare the Protein

Protein preparation is critical. Remove water molecules (unless they mediate key interactions), add missing hydrogen atoms, assign bond orders, and repair missing side chains. Tools like AutoDock Tools, PyMOL, or GNINA built-in preparation can handle this. Set the protonation state relevant to physiological pH (7.4). Save in PDBQT format (AutoDock) or PDB format.

Step 3: Define the Binding Site (Grid Box)

Set a grid box that encompasses the binding site. If the binding site is known (from co-crystallized ligands or mutagenesis data), center the box there. For blind docking (unknown binding site), expand the box to cover the entire protein. Grid box dimensions typically range from 20x20x20 Å (known site) to 60x60x60 Å (blind docking). Grid spacing is usually 1.0 Å (coarse) or 0.375 Å (fine for precise docking).

Step 4: Prepare the Ligand

Obtain the ligand structure as a SMILES string, SDF, or MOL2 file. Sources include PubChem, ChEMBL, or ZINC. Add hydrogens, generate 3D coordinates, set rotatable bonds, and assign Gasteiger partial charges. Save in PDBQT format. For virtual screening of large libraries, use automated ligand preparation scripts.

Step 5: Run Docking

Execute the docking calculation. Popular programs include:

• AutoDock Vina: Fast, open-source, widely used. Runs from command line. Good for screening and production work.
• GNINA: Deep learning-enhanced docking scoring. More accurate than classical scoring functions. GPU-accelerated.
• AutoDock 4: The original. Slower but offers more parameter control. Best for detailed mechanistic studies.
• SwissDock: Web-based. Good for beginners. No command line required.

For AutoDock Vina, the command is: vina --receptor protein.pdbqt --ligand ligand.pdbqt --out output.pdbqt --center_x 0 --center_y 0 --center_z 0 --size_x 20 --size_y 20 --size_z 20

Step 6: Analyze Results

Vina outputs up to 9 binding modes ranked by score. Load the output PDBQT file into PyMOL, ChimeraX, or Discovery Studio. Examine the top-ranked pose: does it fit the binding site? Form hydrogen bonds? Buried in a hydrophobic pocket? Check the binding energy — scores below −7 kcal/mol suggest strong binding. Visualize protein-ligand interactions: hydrogen bonds (typically 2-4), hydrophobic contacts, pi-stacking, and salt bridges.

Docking Scoring Functions Compared

Key Applications of Molecular Docking

Virtual Screening

Docking is used to screen millions of compounds from commercial libraries (ZINC, Enamine, ChemBridge) against a target protein. The top-ranked compounds are selected for experimental testing, dramatically reducing the number of compounds that need to be physically tested. A typical virtual screening campaign tests 10-100 compounds from an in silico screen of 1-10 million.

Binding Mode Prediction

When a new active compound is discovered, docking predicts how it binds to the target — which residues it contacts, what interactions stabilize binding, and whether binding is competitive with natural substrates. This information guides medicinal chemistry optimization.

Structure-Based Lead Optimization

Docking guides the iterative improvement of lead compounds. Chemists modify the ligand structure based on docking predictions, re-dock, and prioritize compounds with improved predicted binding. This cycle reduces synthesis and testing by 5-10x compared to random modification.

Drug Repurposing

Existing approved drugs are docked against novel targets to identify potential new indications. This approach gained prominence during COVID-19, where approved drugs were rapidly tested against SARS-CoV-2 proteins through docking-based virtual screening.

Common Mistakes and How to Avoid Them

Poor protein preparation: Missing hydrogens, incorrect protonation states, and unresolved side chains lead to unreliable results. Always use proper preparation protocols.

Incorrect binding site: Docking to the wrong pocket produces meaningless results. Validate binding site location using literature, co-crystal structures, or computational pocket detection (FPocket, P2Rank).

Over-interpreting scores: Docking scores are approximate. A difference of 0.5 kcal/mol between two compounds is not significant. Always use consensus scoring from multiple programs.

Ignoring protein flexibility: Many targets undergo induced fit upon ligand binding. Use ensemble docking (multiple protein conformations) or flexible receptor protocols.

No experimental validation: Docking predictions must be validated experimentally. Common validation methods include surface plasmon resonance (SPR), isothermal titration calorimetry (ITC), fluorescence polarization, and X-ray crystallography.

Molecular Docking Software Guide