Fig. 5

The TME subtype was an independent prognosis factor in LUAD, SKCM and HNSC. a Univariate Cox regression of the 4 prognostic features (age, gender, Pathologic stage and TME subtype) in TCGA LUAD. b Univariate Cox regression of the 4 prognostic features (age, gender, pathologic stage and TME subtype) in TCGA SKCM. c Univariate Cox regression of the 4 prognostic features (age, gender, pathologic stage and TME subtype) in TCGA HNSC. d Multivariate Cox regression of prognostic features in TCGA LUAD. e Multivariate Cox regression of prognostic features in TCGA SKCM. f Multivariate Cox regression of prognostic features in TCGA HNSC. g DCA of 3-year prediction in TCGA LUAD. h DCA of 3-year prediction in TCGA SKCM. i DCA of 3-year prediction in TCGA HNSC. j The Kaplan–Meier curves of the 4 subtypes in the training and validation cohorts, including TCGA cohorts, GSE68465 (LUAD, n = 442), GSE68571 (LUAD, n = 86), GSE50081 (NSCLC, n = 172), GSE37745 (NSCLC, n = 196), GSE11969 (NSCLC, n = 149), GSE65904 (SKCM, n = 214). DCA, decision curve analysis; LUAD, lung adenocarcinoma; SKCM, skin cutaneous melanoma; HNSC, head and neck squamous cell carcinoma; TME, tumor microenvironment; HH, immune^high stromal^high; HL, immune^high stromal^low; LH, immune^low stromal^high; LL, immune^low stromal^low