Nature Biotechnology: Deep Peptide Recognition Profiling Decodes TCR Specificity for HLA-B27 Ankylosing Spondylitis Autoantigen Discovery
A Nature Biotechnology paper from a multi-institution collaboration introduced deep peptide recognition profiling (PRP) — a yeast-display platform integrated with protein-language models that maps T-cell receptor binding across the proteome for individual TCRs. Applied to HLA-B*27:05-restricted TCRs from patients with ankylosing spondylitis and acute anterior uveitis, the platform identified hundreds to over 6,000 unique peptide ligands per TCR and surfaced candidate autoantigens. Predicted binding scores correlated significantly with experimental T-cell activation across multiple TCRs. The model outperformed AlphaFold3 and tFold-TCR in predicting T-cell activation — meaningful for autoimmune-disease antigen discovery and personalized neoantigen cancer-vaccine design alike. The mechanistic implication: the CDR3β loop arches over the center of the HLA-B27-bound peptide and accounts for most contacts, with CDR3α positioned peripherally.