Fig. 6

The diagnostic model incorporating the A–O panel and tumor purity. A Oligodendroglioma had higher tumor purity than oligodendroglioma (Wilcoxon P = 0.001). B Tumor purity among astrocytoma subtypes based on CNA status (chr 7/chr 10 status, K-W P = 0.690). C Tumor purity among oligodendroglioma subtypes based on CNA status (chr 7/chr 4 loss, K-W P = 0.025). D, E GSVA indicates variation of gene sets enriched in high tumor purity (D) and low tumor purity (E). Blue bars, gene sets significantly enriched in astrocytoma; green bars, gene sets significantly enriched in oligodendroglioma; gray, not significant. F. 962 DEGs selected in high and low tumor purity groups. G–I Confusion matrices of binary results of the diagnostic prediction model in the test sets for AdaBoost (G), GBDT (H) and XGBoost (I). J. Complex ROCs of three machine learning algorithms. K Astrocytoma had significantly higher A–O Panel Classifier (Index) than oligodendroglioma (Wilcoxon ****P < 0.0001). L Validation of the 6 important biomarkers of the A–O panel in the TCGA cohort. M Validation of the 6 important biomarkers of the A–O panel in the CGGA cohort (Wilcoxon test ****P < 0.0001). RF random forest, ROC receiver operating curves, AUC area under the curve, PPV positive predictive value, NPV negative predictive value, OS overall survival, PFI progression-free interval