Document resource
Background Immune checkpoint blockade (ICB) therapy has revolutionized cancer treatment but remains effective only in a subset of patients. A key challenge is the absence of generalizable, pan-cancer biomarkers that accurately capture tumor immune activation and predict response to ICB. Here, we present TIME_ACT, a 66-gene, unsupervised transcriptomic signature of tumor immune activation, derived from melanoma and validated across multiple cancer types and data modalities.Methods TIME_ACT was derived from TCGA melanoma by identifying genes highly upregulated in tumors classified as immune ‘hot’ based on immune score, inflammation score, and spatial TIL density. The signature was then applied across nine additional TCGA cancer types to evaluate its pan-cancer utility. Next, TIME_ACT scores were calculated for 15 independent anti-PD1 transcriptomic cohorts to evaluate their predictive power for treatment response. TIME_ACT scores were further calculated from H&E-stained pathology slides-inferred expression profiles using the deep learning models from Path2Omics, in eight new ICB clinical cohorts.Results TIME_ACT scores robustly identified immune ‘hot’ tumors across cancer types (AUC > 0.96) and positively correlated with IFN-γ signaling, cytotoxic lymphocyte abundance, and immune-rich subtypes. Spatial transcriptomics and histology-based analysis showed that TIME_ACT-high regions are enriched in lymphocyte-dense areas adjacent to tumor cells and exhibit reduced tumor-lymphocyte distances. Next, in 15 anti-PD1-treated cohorts (n = 943, six cancer types), TIME_ACT achieved a mean AUC of 0.76 and a mean odds ratio (OR) of 6.11, outperforming 22 established immune signatures and transcriptomic-based methods for ICB response prediction. Extending beyond transcriptomics, TIME_ACT scores calculated based on inferred tumor transcriptomics directly from pathology slides using Path2Omics framework accurately predicted ICB response in eight new and diverse clinical cohorts; including melanoma, NSCLC, breast, renal, bladder, and HNSCC; with a mean AUC of 0.72 and OR of 5.02, highlighting its broad clinical utility and translational potential.Conclusions TIME_ACT is a robust, pan-cancer biomarker of tumor immune activation that predicts ICB response from both transcriptomic and histopathologic data. By leveraging standard pathology slides, TIME_ACT offers a low-cost, timely, and widely accessible approach for advancing precision oncology and facilitating the effective matching of patients to ICB therapy.