Fig. 4

Characterization of riskscore to predict the prognosis of GBM. (A) The risk score of training set (The data was derived from TCGA). The cut-off value was 1.139165, classifying patients as either high risk (n = 87) or low risk (n = 80). (B) Survival distribution of GBM patients. In the training set, mortality increased as risk scores increased. (C) Kaplan–Meier estimates of survival curves. The K-M curve shows survival time and probability, with the top half representing high and low-risk groups, and the bottom half showing the number of samples remaining at different survival times. (D) Time-dependent ROC curves analysis. The x-axis shows FPR (specificity) and the y-axis shows TPR (sensitivity) in a graph. Higher AUC values indicate greater prediction accuracy, with curves closer to the upper left corner being more accurate. AUC values for 1, 2, and 3 years are all > 0.6, indicating high accuracy. (E) Risk model gene heat map. Red indicates high gene expression and blue indicates low gene expression. (F) Survival distribution of GBM patients in the validation set. (G) The risk score of validation set (The data was derived from CGGA). Mortality increased as risk scores increased. The optimal cut-off value of 2.588629 divided the sample into high and low groups, with 59 patients classified as high risk and 78 patients classified as low risk. (H) Kaplan–Meier estimates of survival curves in the validation set. (I) Time-dependent ROC curves analysis of validation set.The AUC at 1, 2, and 3 years was over 0.6, showing effective prognostic diagnosis efficiency. (J) Risk model gene heat map of validation set.