Accelerating Drug Discovery with Machine Learning
UBC / Vancouver Prostrate Centre leveraging machine learning to advance therapeutic discovery and development
Drug discovery is a complex and lengthy process, the number of possible molecules that can be made with just 30 atoms is astronomically large (1060 molecules) — more than the number of protons/neutrons on earth and the time required to compute all possible structures would be 1035 years, assuming a 200 PF supercomputer.
Deep Docking, the augmentation of traditional molecular docking by leveraging machine learning, provides fast prediction of docking scores of
Glide (or any other docking program) and, hence, enables
structure-based virtual screening of billions of purchasable
molecules in a short time.
Leverage deep learning to augment molecular docking (Deep Docking) to screen 40B+ compounds for multiple SARS CoV-2 viral targets. Early application of Deep Docking algorithms enabled the screening of 1.3 billion commercially available compounds against the novel coronavirus virus in one week — a process that would have taken three years using conventional methods.
I had the opportunity to work with the researchers in setting up a machine learning platform to accelerate their deep docking campaign. As a result, more than a hundred novel SARS-CoV-2 Mpro inhibitor scaffolds were identified with minimal or no human intervention. This study yielded a significant number of novel chemotypes, suitable for future medicinal
chemistry optimization, but also yielded one of the most potent Mpro inhibitors that emerged from only molecular docking.
A detailed description of this research can be found in the November issue of Chemical Science (Issue 48, 2021): Automated Discovery of Noncovalent Inhibitors of SARS-CoV-2 Main Protease by Consensus Deep Docking of 40 Billion Small Molecules
