Fig. 4

Construction and verification of the risk score signature. (a) The univariate Cox regression analysis of the DEGs in TCGA dataset (p-value < 0.05), (b) The multivariate Cox regression analysis of the genes after univariate Cox regression analysis screening in TCGA dataset (stepwise regression), getting the eight genes, such as CENPA, CDCA8, EZH2, KIF20A, KPNA2, CCNB1, KIF18B and MCM4, (c) The correlation between the BIRC5 gene and the eight genes, (d) The number of patients who died between the high- and low-risk groups divided by the median of risk score which was calculated from a risk signature constructed by the eight genes in TCGA database, (e) The number of patients who died between the high- and low-risk groups in ICGC database, (f) The OS between the high- and low-risk groups in TCGA database, (g) The OS between the high- and low-risk groups in ICGC database, (h) The ROC curves at 1- and 3-years for prognostic risk scores in TCGA database, (i) The ROC curves at 1- and 3-years for prognostic risk scores in ICGC database, (j) The expression of the 8 genes between the high- and low-risk groups in TCGA database, (k) The expression of the 8 genes between the high- and low-risk groups in ICGC database. The p-value < 0.05 was considered as statistically significant