Unfiltered TCR beta chain calls for 463 cancer samples and 587 control subjects

Liquid biopsy is a promising non-invasive technology that is capable of diagnosing cancer. However, current ctDNA-based approaches detect only a minority of early-stage disease. We set out to improve the sensitivity of liquid biopsy by harnessing tumor recognition by T cells through the sequencing of the circulating T-cell receptor repertoire. We studied a cohort of 463 patients with lung cancer (86% stage I) and 587 subjects without cancer using gDNA extracted from blood buffy coats. We performed TCR β chain sequencing to yield a median of 113,571 TCR clonotypes per sample and built a TCR sequence similarity graph to cluster clonotypes into TCR repertoire functional units (RFUs). The TCR frequencies of RFUs were tested for association with cancer status and RFUs with a statistically significant association were combined into a cancer score using a support vector machine model. The model was evaluated by 10-fold cross-validation and compared with a ctDNA panel of 237 mutation hotspots in 154 lung cancer driver genes and 17 cancer related protein biomarkers in 85 subjects. We identified 327 cancer- associated TCR RFUs with a false discovery rate (FDR) ≤ 0.1, including 157 enriched in cancer samples and 170 enriched in controls. Levels of 247/327 (76%) RFUs were correlated with the presence of an HLA allele at FDR ≤ 0.1 and tumor-infiltrating lymphocyte TCRs from multiple RFUs bound HLA presented tumor antigen peptides, suggesting antigen recognition as a driver of the cancer-RFU associations found. The RFU cancer score detected nearly 50% of stage I lung cancers at a specificity of 80% and boosted the sensitivity by up to 20 percentage points when added to ctDNA and circulating proteins in a multi- analyte cancer screening test. Overall, we show that circulating TCR repertoire functional unit analysis can complement established analytes to improve liquid biopsy sensitivity for early-stage cancer.<br><br>This dataset contains raw TCR beta chain calls for the N = 463 cancer subjects and N = 587 controls dataset. The columns are as follows.<code>v_gene</code>: V gene of each TCR clonotype<code>j_gene</code>: J gene of each TCR clonotype<code>cdr3_nt</code>: Nucleotide sequence over the CDR3 region<code>cdr3</code>: Amino acid sequence over the CDR3 region<code>templates</code>: Number of UMIs supporting the clonotype<code>sample_name</code>: Sample that the clonotype derived from.The file is in feather format. See the following instructions for reading this file using R (link) or Python (link).<br><br>For details on how the data was generated, please see Li Y. et al. 2025: Circulating T-cell Receptor Repertoire for Cancer Early Detection.