Back

TSTScope Unifies Single-Cell Multi-Omics to Identify Functional T Cell States Predictive of Immunotherapy Response

Cao, S.; Cheng, J.; Wang, F.; Yi, C.; Chen, J.; Wang, K.; Liu, L.; Liu, J.; Li, Y.

2026-01-08 bioinformatics
10.64898/2026.01.08.698283 bioRxiv
Show abstract

Immune checkpoint blockade (ICB) can produce durable responses in cancer, but reliable predictors of benefit are still lacking. CD8 tumor-specific T cells are essential for ICB efficacy, yet it remains unclear which functional states of these cells determine therapeutic success. To address this, we developed TSTScope, an interpretable deep learning framework that integrates single-cell transcriptomic and T-cell receptor sequencing data to generate unified representations of CD8 T-cell identity. By applying TSTScope to non-small cell lung cancer (NSCLC) datasets, we characterized the gene programs defining tumor specificity and computationally inferred a population of potential TSTs. Crucially, we demonstrate that clinical response is not a product of TST abundance, but is instead governed by their functional state. We derived the MPR score, a metric capturing this functional potential, which proved to be a robust predictor of treatment outcomes. In an independent validation cohort, the MPR score significantly outperformed established biomarkers. Collectively, TSTScope identifies a distinct functional state of tumor-specific T cells as a primary determinant of ICB efficacy, providing both a mechanistic framework and a potent tool for precision immunotherapy.

Published in Advanced Science (predicted rank #12) · training set

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.