Deep generative models predict SARS-CoV-2 Spike infectivity and foreshadow neutralizing antibody escape

Abstract

AbstractRecurrent waves of SARS-CoV-2 infection, driven by the periodic emergence of new viral variants, highlight the need for vaccines and therapeutics that remain effective against future strains. Yet, our ability to proactively evaluate such therapeutics is limited to assessing their effectiveness against previous or circulating variants, which may differ significantly in their antibody escape from future viral evolution. To address this challenge, we developed deep learning methods to predict the effect of mutations on fitness and escape from neutralizing antibodies and used this information to engineer a set of 68 unique SARS-CoV-2 Spike proteins. The designed constructs, which incorporated novel combinations of up to 46 mutations relative to the ancestral strain, were infectious and evaded neutralization by nine well-characterized panels of human polyclonal anti-SARS-CoV-2 immune sera. Designed constructs on previous SARS-CoV-2 strains anticipated the antibody neutralization escape of variants seen subsequently during the COVID-19 pandemic. We demonstrate that designed Spike constructs using data available at the time of the implementation of the 2022 bivalent mRNA booster vaccine foretold the level of neutralizing antibody escape observed in the most recently emerging variants. Our approach provides extensive datasets of antigenically diverse escape variants to evaluate the protective ability of vaccines and therapeutics to inhibit future variants. This approach is generalizable to other viral pathogens.