Future-proof vaccine design with a generative model of antibody cross-reactivity
Track: Machine learning: computational method and/or computational results
Cell: I do not want my work to be considered for Cell Systems
Keywords: vaccine design, viral evolution, antibody, protein fitness models, protein language models
TL;DR: We develop a future-proof vaccine design method using a generative model of antibody escape to focus antibody responses to conserved, neutralizing regions unlikely to mutate.
Abstract: Mosaic nanoparticle vaccines incorporating naturally diverse sarbecovirus receptor binding domains (RBDs) represent a promising approach for pan-coronavirus vaccine design. Mosaic nanoparticles elicit broad, cross-reactive immune responses, likely because antibodies utilize avidity effects to preferentially bind to conserved regions where they can cross-link across neighboring RBDs. However, the diversity in natural RBDs is limited, leading to ‘off-target’ antibodies that do not bind to low-mutability regions. We therefore develop a novel future-proof vaccine design method, building upon a probabilistic generative model of antibody escape, to computationally design RBDs with further diversity. This approach aims to focus antibody responses to regions that are (1) neutralizing, (2) accessible and (3) unlikely to mutate during future viral evolution. The designs will be assessed by immunizing mice and testing the breadth of neutralizability of the sera compared to a nanoparticle composed of naturally diverse strains.
Submission Number: 85
Loading