DiffDock is a molecular docking model developed by researchers at the MIT Jameel Clinic that has the potential to one day discover new drugs faster than traditional methods drug discovery methods, while reducing the risk of adverse side effects. Research was led by MIT Jameel Clinic AI faculty lead Regina Barzilay and 2022 MIT Jameel Clinic principal investigator Tommi Jaakkola alongside their PhD students Gabriele Corso, Hannes Stärk and Bowen Jing, who together co-authored a paper introducing the new model at the '11th International Conference on Learning Representations'.

Molecular docking is a method used to model the interaction between two molecules, typically a protein and a ligand (a smaller molecule or ion that binds to the protein). Models like DiffDock generally work by employing computational algorithms to explore the configurations and orientations of a ligand as it interacts with the target protein. The objective is to find the most energetically favourable configuration, which often corresponds to the most likely biological interaction.

In the case of DiffDock, after being trained on a variety of ligand and protein poses, the model is able to successfully identify multiple binding sites on proteins that it has never encountered before. Instead of generating new image data, it generates new 3D coordinates that help the ligand find potential angles that would allow it to fit into the protein pocket. This 'blind docking' approach is significantly more accurate than traditional docking approaches, thanks to its ability to reason at a higher scale and implicitly model some of the protein flexibility.

DiffDock's unique approach to computational drug design represents a paradigm shift from current state-of-the-art tools used by most pharmaceutical companies, revealing a significant opportunity to revolutionise the traditional drug development pipeline.