Docking

​

Protein-Protein Docking

We’ve found that structure prediction using one of our Alphafold3 reproductions is often the best way to do protein-protein docking currently. For Chai, you can also incorporate templates for your known structures. We support both finding templates from the pdb100 or uploading your own custom cif files. We do also offer ColabDock, which is a protein-protein docking tool with the option for specifying known experimental restraints.

​

Protein-Small Molecule Docking

If you need fast docking, we recommend physics based approaches like AutoDock/Smina/Gnina. The output binding affinity from these tools is meant to be an approximation of the binding free energy (ΔG) between the ligand and the receptor, typically reported in kcal/mol. More negative values suggest stronger predicted binding. To use AutoDock/Smina/Gnina, select a binding box on the protein for where you know your ligand to be by specifying its x/y/z and height/width/depth (can be the whole protein if you have no idea). If you don’t require fast runtime, we recommend using Chai or Boltz, which are structure prediction tools which have also been found to perform well at docking tasks. For example, Chai-1 scores 77% on the posebusters benchmark set, compared to 60% for Autodock-Vina. If you have a known binding site, you can specify a pocket restraint between your pocket residue and ligand chain. Reach out to info@tamarind.bio if you’re interested in virtual screening against a ligand library. We offer a multi-stage approach, with more computationally intense screening for a subset of ligands which pass initial screening.