Fig. 1

The process of constructing PRGs-related clusters and prognostic signatures. A Unsupervised clustering of differential expression genes between tumor patients and normal person in the TCGA and GTEx cohorts. Heatmap show 16 differential gene expression levels with P < 0.001. Red indicates up-regulation of gene expression, and blue indicates down-regulation. B During clusters k = 2, consensus map of NMF clustering in TCGA training cohort. C The Kaplan-Meier curves for the two clusters based on TCGA cohort. D The hazard ratios (HR) and 95% confidence intervals (CI) calculated by univariate Cox regression are shown. E Each independent variable’s changing trajectory. A horizontal axis represents the logarithm of lambda, and a vertical axis represents its coefficient. Increasing lambda increases the number of independent variable coefficients that tend to 0. Confidence intervals for each lambda. When the lambda is 0.02748, an optimal model is acquired. F The scattergrams of the RiskScore value (up) and survival status (down) of LGG patients in the TCGA cohorts. G-J The 1-, 3- or 5-year OS predicted ROC curves of a prognostic signature in the TCGA, CGGA, mRNA array 301, and Rembrandt cohorts. The 1-, 3- or 5-year AUC is shown in the Figure.