Integrating computational and experimental approaches to identify antigen–specific antibodies

The identification of therapeutic antibodies is critical for developing effective treatments for many diseases. We present a joint computational and experimental approach for identifying antigen–specific antibodies. We develop and test our pipeline on a probe antigen: GFRAL fused with Diphtheria Toxin Fragment A (DTA). Using humanized Trianni mice, we conducted a longitudinal study with repeated blood sampling and splenic analysis. We applied the STAR computational method that exploits similarity network properties on bulk antibody repertoire data sampled at key time points. By mapping the output from STAR to single cell data taken at the last time point, we successfully identified a pool of antibodies, of which 45% demonstrated binding capabilities. We observed convergent selection, where responding sequences with identical amino acid complementarity determining regions 3 (CDR3) were found in different mice. We provide a catalog of 70 experimentally validated antibodies against the GFRAL-DTA fusion antigen. This study underscores the utility of integrating computational methods and experimental validation for antibody discovery in therapeutic contexts by reducing time and increasing efficiency.