Back

Pan-cancer prediction of tumor immune activation and response to immune checkpoint blockade from tumor transcriptomics and histopathology

Mukherjee, S.; Patiyal, S.; Pal, L. R.; Chang, T.; Biswas, S.; Dhruba, S. R.; Stemmer, A.; Singh, A.; Yousefi-Rad, A.; Chen, T.-H.; Wang, B.; Marino, D.; Shon, W.; Yuan, Y.; Faries, M.; Hamid, O.; Reckamp, K.; Waissengrin, B.; Ornelas, B.; Chu, P.-Y.; Boudjadi, S.; Ley, L.; Akbulut, D.; Ahmar, N. E.; Signoretti, S.; Braun, D. A.; Joo, H.; Kim, H.; Osipov, A.; Figlin, R. A.; Bar, J.; Barshack, I.; Day, C.-P.; Sargsyan, K.; Apolo, A. B.; Aldape, K.; Yang, M.-H.; Atkins, M. B.; Ronai, Z. A.; Hoang, D.-T.; Ruppin, E.

2025-06-30 cancer biology
10.1101/2025.06.27.661875 bioRxiv
Show abstract

Accurately predicting which patients will respond to immune checkpoint blockade (ICB) remains a major challenge. Here, we present TIME_ACT, an unsupervised 66-gene transcriptomic signature of tumor immune activation derived from TCGA melanoma data. First, TIME_ACT scores accurately identify tumors with activated immune microenvironments across cancer types. Analysis of spatial features of the tumor microenvironment revealed that TIME_ACT-high regions exhibit dense lymphocyte infiltration near tumor cells, indicating localized immune activation. Second, in 15 anti-PD1 transcriptomic cohorts spanning six cancer types, TIME_ACT outperforms 22 established signatures and methods, achieving a mean AUC of 0.76 and a clinically meaningful mean odds ratio of 6.11. Thirdly, TIME_ACT scores can be accurately inferred from tumor histopathology slides. Finally, slide-inferred TIME_ACT scores predict ICB response across eight unseen cohorts, achieving a mean AUC of 0.72 and a mean odds ratio of 5.02. These findings establish TIME_ACT as a robust, pan-cancer, and low-cost predictor of ICB response.

Matching journals

The top 4 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.