Figure 3. Internal and external validation of MAGs to determine its clinical predictive value. (A, C) The AUCs of ROC curves were 0.763 and 0.803 in predicting 3-year OS events in training and testing cohorts, respectively. (B, D) Besides, Kaplan-Meier analysis indicated that patients with high MAGs-score suffered significantly worse OS outcomes (P = 2.904e-08), which was validated consistently in testing cohort with P = 1.031e-10. (E, F) In addition, we also proved our findings in an independent ICGC cohort and observed the similar statistical results. (G–I) We further integrated MAGs signature with survival analysis in the total TCGA-KIRC cohort and distribution plots suggested that high MAGs risk scores correlated with more dead and recurrence/progression cases.