A reproducible pipeline for TCR β-chain repertoire analysis from thoracic tumour RNA-seq and ICI-associated cardiac immunotoxicity.

BACKGROUND: Bulk RNA-sequencing of thoracic tumours is routine, but the T-cell receptor (TCR) repertoires it contains are rarely re-mined with a uniform pipeline, and library-prep compatibility for productive V-J-spanning assembly is rarely audited at the cohort level for non-small-cell lung cancer (NSCLC) and oesophageal squamous-cell carcinoma (ESCC). OBJECTIVE: We built a reproducible, openly-released pipeline for productive T-cell-receptor β-chain (TRB) repertoire reconstruction from public thoracic-tumour RNA-seq, and extended it to immune-checkpoint-inhibitor (ICI)-associated cardiac immunotoxicity. METHODS: Four GEO cohorts (NSCLC and ESCC bulk RNA-seq) were assembled with TRUST4 under a transparent cohort-level QC rule (≥ 10% V+J assemblies on ≥ 1 representative sample), and antigens were annotated by exact CDR3-amino-acid matching to VDJdb. The identical pipeline was applied to 23 peripheral-blood TRB repertoires from an ICI-irAE single-cell cohort (GSE180045; 10 discovery and 13 validation libraries). RESULTS: The retained pilot dataset comprised 12 productive samples and 916 productive TRB clones; ESCC (n = 1 after QC) was oligoclonal versus diverse NSCLC. Of 37 VDJdb hits, 33 were viral and 4 tumour or self, with no clone shared across cohorts. Thoracic-tumour tissue repertoires were significantly more oligoclonal than peripheral ICI-irAE repertoires (top-10 clone fraction 0.76 vs 0.18; inverse Simpson 8.8 vs 177; Mann-Whitney BH-FDR < 0.001), whereas depth-normalised metrics did not differ (FDR = 0.61), indicating an assay-platform (bulk RNA-seq versus 10x single-cell), sequencing-depth and tissue-versus-blood effect rather than intrinsic clonal architecture; within-patient cardiac contrasts were not significant and are reported descriptively. CONCLUSION: One uniform pipeline transparently re-mines public thoracic-tumour and cardiac-immunotoxicity RNA-seq for TCR repertoires. These findings are descriptive and hypothesis-generating rather than statistically generalisable, with the ESCC-versus-NSCLC contrast resting on a single ESCC sample.