Molecular Docking
Molecular docking predicts how two molecules fit together in 3D and a staple method in computer-aided drug design. For example, a small molecule drug with its target enzyme, an antibody drug with its target antigen, or a peptide hormone with its natural receptor. Docking is also called cofolding when folding (i.e. protein structure prediction) is done at the same time. By studying how molecules dock into a complex, scientists can learn how a drug-target interaction is working as part of the drug discovery process or better understand how our bodies work. While experimental results from cryo-EM or crystallography are the gold standard, docking offers a much faster and economical computational solution.
Dock asparagine into A0A1L9RXX7
Protein-Small-Molecule Docking
DiffDock
DiffDock is a diffusion-based ML model for the docking of a protein to a small molecule. DiffDock reports a confidence score (-inf, inf) where a larger number is better, and scores below -1.5 can be considered low confidence. Note the confidence score does not predict the affinity, aka strength of the biochemical attraction, of the complex.
Inputs
PDB file: specifies the protein structure in 3D.
SMILE string: specifies the identity of the small molecule.
Outputs
PDB file: specifies the predicted structure of the protein + small molecule docked together.
Confidence score: model confidence in the dock
Example Scripts
Load the protein structure for uniprot ID P08100, load the small molecule retinal, and then dock the two together.
Dock biotin into P22629
Protein-Protein Docking
AlphaFold3
AlphaFold3 (aka AF3) is a newer version of F2 using diffusion for predicting the 3D structure of proteins + nucleic acids + small molecules + post-translational modification. It is capable of multi-chains and therefore is also a cofolding or docking method.
AlphaFoldMultimer
AlphaFoldMultimer is a variation of AF2 trained specifically for protein-protein cofolding/docking.
Protein + Nucleic Acid
AlphaFold3
AlphaFold3 (aka AF3) is a newer version of AF2 using diffusion for predicting the 3D structure of proteins + nucleic acids + small molecules + post-translational modification. It is capable of multi-chains and therefore is also a cofolding or docking method.
RoseTTAFold All-Atom
RFAA is a variation of RoseTTAFold for predicting the 3D structure of proteins + nucleic acids + small molecules + post-translational modification. It is capable of multi-chains and therefore is also a cofolding or docking method.
RoseTTAFoldNA
RoseTTAFoldNA is a variation of RoseTTAFold for AI structure prediction of protein + DNA + RNA. Capable of multi-chains and therefore also a cofolding or protein-nucleic acid cocking method.
Analyzing Molecular Docking Results
High quality computational evaluation of docking results is an underdeveloped area.
- Comparison to Experiment: Alignment to and root-mean-square deviation (RMSD) evaluation of the docked structure to experimental data would be the gold-standard metric. However, in the vast majority of cases experimental data is unavailable.
- Visual Inspection: Use of a molecule visualizer with expert knowledge to check if docked structure makes sense with biochemistry or data that is not a full experimental structure. For example, checking if there is a strong bonding pattern and shape complementarity at the interface. Or for example, checking if the interface involves residues known to be important for the interaction, perhaps from mutational scanning experiments.
Connect Copilot with Your Chosen Tools
Binding Site Prediction
Docking presumes the identity of both molecules is known. If only one protein is known, users might be interested in predicting where on the protein other molecules are likely to bind as not all parts of a protein can easily interact with other objects. The places are often called pockets or hotspots.
Folding
Docking can require that the individual input structures (especially of the protein) to be known beforehand (though this is changing quickly). Therefore, AI protein structure prediction models can be used together with docking models.
Affinity Prediction
Docking answers the question of how the atoms in two molecules come together in 3D, but many users will be interested in predicting the biochemical strength of the interaction, or affinity.
- AutoDock Vina: A non-AI based computational method considered the gold standard for affinity between a protein and a small molecule ligand.