Fig. 2

Unsupervised clustering of tumor microenvironment (TME) cell populations in the Cancer Genome Atlas (TCGA) cohort. a Consensus matrix heatmap from K-means consensus clustering identified 3 different clusters—namely, C1, C2, and C3. b Kaplan-Meier curves showed the overall survival rates for patients with hepatocellular carcinoma (HCC) in TCGA and Liver Cancer-RIKEN,JP (LIRI_JP) cohorts. Patients were divided by different TME subtypes. Violin plot of the absolute abundance of TME cell populations distinguished by different TME subtypes in the TCGA cohort (c) and LIRI_JP cohort d. Abbreviation: ns, no significance. *p < 0.05, **p < 0.01, ***p < 0.001, ***p < 0.0001). A log-rank test was used to determine the statistical significance of differences. e Immune and stromal scores were determined using ESTIMATION for patients of 3 TME subtypes in the TCGA cohort. Medians, interquartile ranges, and minimum/maximum are shown in boxplots, and the differences between each group were compared by using the Mann–Whitney U test