Abstact

Antibodies that recognize peptide-loaded class I major histocompatibility complex (MHC) molecules could enable therapeutic targeting of intracellular oncogenic proteins, yet their discovery has been hampered by the small size of peptide antigens and the need for allele-specific recognition. Here, we describe an integrated in silico-in vitro workflow for generating high-affinity, selective antibodies to Kirsten rat sarcoma viral oncogene homolog (KRAS) G12D(10) presented by human leukocyte antigen (HLA)-C *08:02, a clinically validated cancer neoantigen. In the in silico stage, multiple human antibody-derived variable fragments optimally docked to the target were generated, followed by limited sequence design of complementarity-determining regions (CDRs). In the in vitro stage, limited CDR diversity was introduced to construct a yeast surface-displayed library, which was subjected to iterative selection. This workflow yielded KRAS G12D(10)/HLA-C *08:02-specific antibodies with high affinity. Comprehensive specificity profiling using a phage display library confirmed the absence of human off-target reactivity, and in silico analysis predicted low immunogenicity. When reformatted as chimeric antigen receptors or bispecific T cell engagers, these antibodies mediated selective cytotoxicity against target-positive cells. Together, these findings establish a practical design-to-function pipeline for T cell receptor (TCR)-like antibody discovery and demonstrate the feasibility of therapeutically targeting KRAS G12D-driven malignancies.