UcTCRp：仅用 TCRβ 识别 MAIT 与 iNKT 相关免疫组库信号的框架

Mucosal-associated invariant T (MAIT) cells and invariant natural killer T (iNKT) cells are conventionally identified using invariant or semi-invariant TCRα chains, antigen-loaded tetramers, or transcriptomic phenotypes. These requirements limit their detection in public and clinical immune-repertoire datasets that contain only TCRβ sequences. Here we present UcTCRp, a TCRβ-only framework for detecting and profiling relative MAIT- and iNKT-associated signals in bulk immune repertoires. We first identified model-informative TCRβ features, including biased V-gene usage and reproducible CDR3β amino acid composition, in collected human and mouse MAIT and iNKT TCRβ sequences. UcTCRp subsequently integrates these features using species-specific transformer-based representation pretrained separately on large human and mouse TCRβ repertoires, followed by independent supervised modeling with curated MAIT, iNKT, and conventional T cell references for each species. The resulting species-specific models generalized robustly across independent human and mouse cohorts. In paired scRNA-seq/scTCR-seq datasets, UcTCRp recovered transcriptome-defined MAIT and iNKT cells and identified additional MAIT-like candidates supported by receptor evidence but missed by expression-only annotation. Bulk calibration against paired single-cell references and synthetic spike-in experiments established operating characteristics for repertoire-level abundance estimation. These results establish unpaired TCRβ repertoires as an actionable substrate for detecting relative unconventional T cell-associated immune signals, enabling archived repertoire resources to be repurposed for systems-level studies of tissue immunity, disease, and therapeutic response.