Rapid in silico design of antibodies targeting SARS-CoV-2 using machine learning and supercomputing

Thomas Desautels; Adam Zemla; Edmond Lau; Magdalena Franco; Daniel Faissol

In response to the COVID-19 pandemic, we modified and used our computational components of the antigen design platform to propose mutations to SARS-CoV-1 neutralizing antibodies to achieve and optimize binding to the receptor binding domain (RBD) of the SARS-CoV-2 spike protein. This approach is motivated and enabled by knowledge of existing antibodies specific to the RBD of the SARS-CoV-1 spike protein that prevent SARS-CoV-1 from binding the human ACE2 receptor and entering the cell, thus neutralizing the virus [10, 11] . The high similarity of SARS-CoV-1 and SARS-CoV-2, including the RBD [12] , suggest that such an approach could produce efficacious therapeutic antibodies. In 22 days (1/23-2/13/2020), we demonstrated that our platform could generate antibody designs using only SARS-CoV-2 sequence information, with support from antibody/antigen structures that had previously been experimentally determined for SARS-CoV-1. We executed this entire workflow by constructing a homologybased structural model before any experimentally determined structures of the SARS-CoV-2 Sprotein were available.