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

3. Define a set of residues for modification in each of the starting SARS-CoV-1-neutralizing antibodies via automatic contact estimation. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.03.024885 doi: bioRxiv preprint D92_L, S93_L, S94_L, D95A_L, (Y96_L). We did not consider insertions or deletions at this phase of our work. 4 . Use machine learning module of computational design platform to iteratively propose mutations to the original antibody (M396) and perform free energy calculations using FoldX [21] on LLNL high-performance computing (HPC) to maximize estimated affinities to our SARS-CoV-2 spike protein homology model. By Feb 13, 2020, we evaluated 89,263 mutant antibodies selected from a design space of 10 40 (20 amino acids 31 positions ). The primary template for this structure was PDB 2dd8.