Molecular docking

What is molecular docking?

Molecular docking is a computational method for generating and ranking plausible bound configurations of two molecules. The term covers small-molecule docking, macromolecular complex docking, peptide and antibody docking, global searches, flexible and ensemble protocols, and learned pose-prediction models. Docking produces hypotheses rather than direct measurements of binding or biological activity.

A docking method combines a search with a scoring or confidence model. The search explores relative position, orientation, and internal conformation; the model ranks candidate complexes using simplified physical terms, statistical potentials, or learned structural patterns. Protein–ligand methods search a small molecule in a pocket, while macromolecular methods handle larger, more flexible interfaces.

Docking is not one universal calculation. Local docking assumes a known site; blind docking searches broadly; flexible and ensemble methods represent receptor motion; consensus docking compares methods; and AI docking uses learned generative models. Choose the workflow from the molecular partners, available evidence, and uncertainty that the calculation must represent.

When to use molecular docking

• Pose generation. Build testable 3D hypotheses for how molecular partners may interact.
• Comparative modeling. Compare methods, receptor states, pockets, or partner-specific strategies.
• Structure-guided decisions. Prioritize complexes for inspection, refinement, screening, or experimental validation.

Benefits of molecular docking

• Structural hypotheses. Returns explicit complexes that can be inspected and tested.
• Method choice. Supports distinct partner classes and flexibility assumptions.
• Connected follow-up. Links pose generation to interaction, geometry, and stability review.

Primary limitations

• Approximate scoring. Docking scores and model confidence are not experimental affinity.
• Preparation sensitivity. Structures, protonation, chemistry, and search boundaries affect results.
• Biological validation. A plausible pose does not establish interaction or function.

Types of molecular docking

Choose the docking use case that matches the molecular partners and the uncertainty you need to represent.

Protein–ligand docking

Predict small-molecule binding poses in a protein target and inspect scoring and interaction evidence.

Best for: Hit interpretation, pose generation, and structure-guided design
Requires: A prepared protein structure and ligand structure

Protein–protein docking

Predict how two protein partners assemble into a biomolecular complex.

Best for: Interaction modeling and interface hypotheses
Requires: Structures or suitable models for both protein partners

Antibody–antigen docking

Explore antibody recognition orientations with antibody-aware interface evidence.

Best for: Epitope and paratope hypotheses
Requires: Antibody and antigen structures or defensible models

Peptide–protein docking

Model a flexible peptide partner against a protein receptor.

Best for: Peptide binders, motifs, and signaling interfaces
Requires: A receptor structure and peptide structure or model

Blind docking

Search a protein broadly when the relevant binding site is unknown.

Best for: Binding-site hypothesis generation
Requires: A prepared target structure and ligand

Flexible molecular docking

Account for ligand flexibility and selected or learned receptor movement.

Best for: Targets with induced fit or uncertain side-chain states
Requires: A prepared target, ligand, and a justified flexibility model

Ensemble docking

Dock against multiple receptor conformations instead of one static structure.

Best for: Conformationally heterogeneous binding sites
Requires: A curated receptor ensemble and consistent preparation

Consensus docking

Compare poses and rankings from multiple docking methods.

Best for: Robustness checks and method-disagreement review
Requires: Consistent inputs and a predefined comparison rule

AI molecular docking

Use learned diffusion or flow models to predict protein–ligand complexes.

Best for: Fast pose hypotheses and complementary model comparison
Requires: Compatible protein and ligand inputs

What a molecular docking calculation actually does

Docking starts from structural representations of two molecular partners. Depending on the method, one partner may remain fixed while the other translates, rotates, and changes torsional angles, or both partners may undergo limited or learned conformational adjustment. The algorithm samples only a fraction of the possible configurations, so search settings and starting structures shape what solutions can be found.

The resulting score is an approximation designed to rank candidate configurations efficiently. AutoDock Vina reports a Vina-family energy estimate, GNINA adds learned CNN signals, macromolecular methods use their own weighted terms, and generative models return model-specific confidence. These values are not interchangeable and none is a direct experimental measurement of affinity, kinetics, occupancy, or biological effect.

• Search space. Defines where the partners may move and which ligand, side-chain, backbone, or rigid-body degrees of freedom are sampled.
• Pose generation. Produces alternative candidate complexes rather than one uniquely determined structure.
• Ranking and review. Orders candidates with a method-specific score or confidence that must be interpreted alongside geometry and independent evidence.

How to choose a molecular docking method

Start with the molecular partners. Protein–ligand methods do not automatically transfer to protein–protein, antibody–antigen, or peptide–protein complexes because the search spaces and scoring assumptions differ.

Then decide whether the site is known, how receptor flexibility should be represented, and whether multiple conformations or methods are needed. Preserve each method’s native output and validate the protocol on relevant known systems whenever possible.

• Known pocket and small molecule. Start with protein–ligand docking and a local search region supported by a co-crystal ligand, residues, or pocket evidence.
• Unknown pocket. Use blind docking for binding-site hypothesis generation, ideally with independent cavity detection and explicit false-positive review.
• Protein or peptide partner. Choose a macromolecular method designed for protein–protein, antibody–antigen, or peptide–protein interfaces.
• Conformational uncertainty. Use flexible or ensemble docking when structural evidence shows that one rigid receptor state is inadequate.
• Method uncertainty. Use consensus or AI-method comparison to expose sensitivity to scoring and learned-model assumptions.

How to do molecular docking online

ProteinIQ provides runnable templates for each maintained docking use case. The exact inputs and tools change with the partner class, but the preparation and review principles remain consistent.

1. Choose the docking type. Select the use case that matches the molecular partners, binding-site knowledge, receptor flexibility, and intended decision.
2. Prepare every structure. Check chains, missing atoms, cofactors, waters, protonation, ligand stereochemistry, charge, and the biological state represented by each input.
3. Open the matched workflow. Use the page workflow template so the selected tools, input ports, and method comparison reflect that docking problem.
4. Configure the scientific boundary. Define the pocket, restraints, flexible residues, receptor conformations, model settings, or comparison rule before running.
5. Inspect and export all evidence. Review poses, complex geometry, native scores, confidence, logs, failures, and provenance, then download structures for validation and reporting.

How a molecular docking workflow works

A defensible workflow prepares inputs consistently, chooses a scientifically matched search strategy, runs one or more methods, and reviews geometry and uncertainty before validation.

1. Classify the system. Identify the partner class and biological question.
2. Prepare structures. Repair structures and standardize ligand chemistry.
3. Define the search. Set site, restraint, flexibility, or ensemble boundaries.
4. Generate models. Run complementary methods and preserve native outputs.
5. Validate. Inspect geometry and compare with orthogonal evidence.

Inputs and outputs

Check formats before running, then inspect and download the result from every workflow step.

Inputs

• Molecular structures. PDB SDF SMILES Prepared structures appropriate to the selected partner class.
• Search evidence. Binding sites, restraints, reference complexes, receptor ensembles, or known interactions when available.

Outputs

• Candidate complexes. PDB PDBQT SDF Method-native ranked poses or complex models.
• Review package. Scores, confidence, contacts, logs, failures, provenance, and downloadable files.