Fig. 4: Enrichment of functional states has prognostic value in cancer patients.

a Heatmap showing bulk RNA sequencing gene expression levels of differentially expressed genes for different T cells subsets in post-tx 4T1 tumors. Each raw represents an individual mouse (n = 4 biologically independent mice for control, CTLA4, and RT groups, n = 6 biologically independent mice for RT + CTLA4). Ctla4i and RT indicate CTLA-4 checkpoint blockade and radiation therapy, respectively. Green boxes below heatmap indicate that genes were statistically significantly differentially expressed for a specific functional state in the single-cell sequencing data. Blue boxes indicate genes used in the combined gene signature in (d–g). b Univariate and (c) multivariate cox proportional hazard regression models were used to assess the association between gene enrichment score and survival in patients with triple-negative breast cancer (TNBC) using the METABRIC dataset (n = 299 biologically independent samples). d Multivariate cox proportional hazard regression model and (e) Kaplan–Meier curves of survival of patients with TNBC from the METABRIC dataset by enrichment of a combined gene signature that incorporated differentially expressed genes from CD8 EA/C1, CD8 EM/C1, and CD8 Tpex/C6. Difference in survival between the two groups was assessed using the log rank test. f Multivariate cox proportional hazard regression model and (g) Kaplan–Meier curves of survival of patients with skin cutaneous melanoma from TCGA SKCM dataset (n = 417 biologically independent samples) and enrichment of a combined gene signature that incorporated differentially expressed genes from CD8 EA/C1, CD8 EM/C1, and CD8 Tpex/C6, p-value was based on the log rank test. NPI Nottingham Prognostic Index, CD3E expression of CD3E in tumors, AJCC American Joint Committee on Cancer. Bars represent the 95% confidence interval for the hazard ratio (b–d, f). Source data are provided as a Source Data file.