Integrating Atomic-Scale Antigen Modeling and Repertoire-Scale Protein Language Models
The field of Immunology still struggles to connect what a single antibody does with how a person’s whole antibody repertoire protects as viruses change.
During COVID-19 we often spotted escape variants only after they spread widely. Lab measurements of circulating binding and neutralization antibodies were identified as correlates of protection against contemporaneous viruses. However, with rapidly and unpredictable evolution of viruses, we need innovative methods to examine the full array of antibodies and their relationships within an individual and in a population to translate into real-world protection. Although we can now sequence B-cell receptors and catalog repertoires, we lack tools that predict which variants a repertoire can cover over time. RP1 closes this gap by combining molecular dynamics of binding and kinetics with antibody language models, linking molecular interactions to repertoire behavior. The research will focus on HIV-1 and SARS-CoV-2 to capitalize on the extensive existing data sets for these pathogens. The goal is a practical framework that forecasts vulnerability and protection as pathogens evolve and supports more durable biomedical interventions.