A Generative Deep Learning Approach for the Discovery of SARS CoV2 Protease Inhibitors

COVID19 has caused thousands of deaths worldwide within a few months. The rapid spread of this virus that causes COVID19, termed SARS CoV2, has been facilitated by the lack of effective vaccines and treatments against this virus. In recent months, our team has developed a novel deep learning platform, Rosalind, for drug design and optimisation, and it enables rapid in silico discovery and evaluation of novel chemical designs. In the current work, we applied Rosalind for the rapid discovery of SARS CoV2 replication inhibitors that target the virus main protease M^pro. Through a series of training and optimisation rounds based on reported SARS CoV2 M^pro inhibitors helped by docking into the recently reported crystal structures of SARS CoV2 M^pro and medicinal chemistry input, we identified the a series of promising SARS CoV2 M^pro inhibitors. These compounds are presented in this work so they scientific community could pursue them while we continue our deep learning-based work in a collaborative manner to identify lead SARS CoV2 M^pro compounds with excellent drug-like properties that could be developed in a timely manner to address the urgent need for new and effective COVID19 treatments.