Fig. 2

Prognostic significance of JAK-STAT driver genes. a Heatmap illustrates hazard ratio values obtained from Cox proportional hazards regression on 40 candidate drivers across all cancer types. b Heatmap illustrates Spearman’s correlation coefficient values comparing hazard ratios of the 40 driver genes. Highly correlated genes are highlighted in red and are demarcated by a red box. c Box plots represent the distribution of 28-gene scores derived from highly correlated JAK-STAT driver genes. Cancers are ranked from high to low median scores. d Heatmap depicts the Z-scores for each of the 28 genes by cancer types. e Kaplan-Meier analyses confirmed prognosis of the 28-gene signature in five cancer cohorts. Patients are separated into 4th and 1st survival quartiles based on their 28-gene scores. P values are obtained from log-rank tests. f Ordination plots of multidimensional scaling analyses using signature genes reveal significant differences between tumor and non-tumor samples. P values are obtained from PERMANOVA tests. g Expression distribution of 28-gene scores in tumors stratified by stage (s1, s2, s3 and s4). P values are determined using ANOVA