Dissecting serum polyclonal antibody escape to SARS-CoV-2 variants by deep mutational learning.

Deep mutational scanning (DMS) has been extensively used to investigate how single-position mutations in the receptor-binding domain (RBD) affect binding of ACE2 and neutralizing antibodies, thus revealing mutations that drive immune escape. However, these mutations represent only a small fraction of the combinatorial mutational space. Here, we present deep mutational learning (DML), a method integrating combinatorial mutagenesis, yeast surface display, deep sequencing, and machine learning (ML), to analyze serum polyclonal antibody escape. Sera from coronavirus disease 2019 (COVID-19)-vaccinated individuals were screened against diverse RBD variant libraries, validating 300 serum-variant interactions across 10 individuals. Performance remained consistent across ML architectures, suggesting that serum binding and escape are governed by distinct, localized RBD sequence features. Notably, escape profiles were highly individualized, whereas binding signatures were more conserved, reflecting convergent epitope targeting. Thus, DML can be used to generalize beyond observed RBD variants, assess cohort-specific immune breadth, and inform vaccine design in the face of viral evolution.