Disease associated human TCR characterization by deep-learning framework TCR-DeepInsight

Abstract

SUMMARYT cell function is defined by both T cell receptors (TCR) and T cell gene expression (GEX). Although single-cell technology enables the simultaneous capture of TCR and GEX information, the lack of a reference atlas and computational tools hinders our ability to uncover the fundamental TCR usage rules and to efficiently characterize disease-associated TCRs (dTCR). Here, through the collection of million-scale single-cell GEX-TCR reference atlas comprising 20 diverse disease conditions, we revealed the intrinsic features of TCR-MHC (Major Histocompatibility Complex) restriction in CD4/CD8 lineages. We observed the higher coherence for TCRα/β chains in memory T cells, and detected widely-existing public TCRα/β pairs across individuals. Building upon the reference atlas, we introduced TCR- DeepInsight, a deep-learning framework featuring a disease specificity scoring system that enables the characterization of dTCR clusters with similar GEX-TCR. Our study provides a valuable tool for researchers to analyze single-cell GEX-TCR data and identify dTCRs comprehensively and robustly.