Fig. 4: ECM signature for prognostic prediction in IDH-mutant gliomas.

a Partial likelihood deviance plot showing the optimal selection of ECM signature genes. b Distribution of risk scores and survival times for each patient, with high-risk patients shown in red and low-risk patients in blue. The expression levels of key genes (SULF1, CSPG5, COL11A1, and CLCF1) are also shown for each patient, with higher expression levels associated with the high-risk group. c Kaplan–Meier survival analysis comparing OS between high-risk and low-risk groups, with a significant difference (P < 0.0001) in survival probability. d ROC curves showing the predictive performance of age, gender, grade, and risk score for OS, with risk score exhibiting the highest AUC (AUC = 0.854). e ROC curves at 1, 3, and 5 years for risk score, demonstrating the robustness of the ECM signature in predicting long-term survival outcomes. f Concordance index (C-index) for risk score, age, gender, and grade over time, with risk score showing superior predictive accuracy. g Nomogram combining risk score, age, gender, and grade for predicting OS at 1, 3, and 5 years. The corresponding survival probabilities for each factor are displayed, with risk score and grade having the greatest impact on OS prediction. h Calibration plot comparing nomogram-predicted OS with observed OS at 1, 3, and 5 years, showing excellent agreement between predicted and observed outcomes.