Protein–ligand docking is the archetypal computer-assisted drug design workflow. It directly addresses the key questions of "how will this compound bind to my target" and "will it be any good," making it immediately relevant to almost every drug discovery campaign.
Unfortunately, docking also doesn't work very well.
Traditional docking tools generate binding poses that often look reasonable but can be physically impossible. Among other difficulties, many methods struggle to account for ligand strain, leading to false positives that waste valuable synthesis time and resources.
So why bother with docking at all? Because when done right, it remains one of the most valuable tools in a medicinal chemist's arsenal.
We've built a docking tool that combines the tried-and-true AutoDock Vina method with a fast ligand-strain-energy calculation.
AutoDock Vina is one of the most widely used and well-known docking methods. Rowan offers access to Vina through an integrated interface that makes it easy to submit jobs, run dozens of docking runs in parallel, and quickly view the results—all without having to touch AWS, format an input file, or download any storage-hogging desktop applications.
By running a conformer search on each ligand before docking it with AutoDock Vina, Rowan can achieve results on par with the most advanced neural networks while maintaining optimal speed, computational cost, and generalizability.
To help improve the robustness of docking scores, Rowan calculates a strain energy for each docked pose. We do this by finding the lowest energy conformers of each ligand and comparing them to the docked poses of that ligand.
Our approach compensates for minor bond length variations and other artifacts introduced by Vina by employing AIMNet2 optimization with harmonic constraints on heavy atoms, resulting in a measurement that accurately captures true angle and dihedral strain energy while preserving the essential binding pose.
This strain metric can be used to discard unphysical poses, increasing the effectiveness and generalizability of large-scale docking.
Once a docking run is complete, results are shown in Rowan's custom viewer.
By default, the protein is shown in ribbon-diagram form. Residues that interact with a docked ligand are shown in all-atom view, and there are settings to control whether or not to display nearby non-interacting residues and much more.
The viewer has modes to focus on the selected docked pose, the pocket, and the entire protein. All docked poses can also be displayed simultaneously, providing a high-level visual overview of the run.
