Leveraging machine learning for integrative analysis of T-cell receptor repertoires in colorectal cancer: Insights into MAIT cell dynamics and risk assessment

Abstract

This study investigates the T-cell receptor (TCR) repertoires in colorectal cancer (CRC) patients by analyzing three distinct datasets: one bulk sequencing dataset of 205 patients with various tumor stages, all newly diagnosed at Sheba Medical Center between 2017 and 2022, with minimal recruitment in 2014 and 2016, and two (public) single-cell sequencing datasets of 10 and 12 patients. Despite the significant variability in the TCR repertoire and the low likelihood of sequence overlap, our analysis reveals an interesting set of TCR sequences across these data. Notably, we observe elevated presence of mucosal-associated invariant T (MAIT) cells in both metastatic and non-metastatic patients. Furthermore, we identify nine identical TCR alpha and TCR beta pairs that appear in both single-cell datasets, with 13 out of 18 sequences from these sequences also appearing in the bulk data. Clinical risk analysis over the bulk dataset, using a subset of these unique sequences, demonstrates a correlation between TCR repertoire disease stage and risk. These findings enhance our understanding of the TCR landscape in CRC and underscore the potential of TCR sequences as biomarkers for disease outcome.