Gene selection and supervised model classifications on TCGA Glioma dataset. (A) Alluvial diagram showing the consistencies and discrepancies between different classifications methods for TCGA Glioma dataset. Colors represent the IDH-codeletion status of the samples. *indicates the IDHwt Astrocytoma cases †indicates the IDHmut Glioblastoma cases. (B) Expression heatmap of TCGA Glioma cases (columns) for 168 selected genes (rows). Sample categories are shown on the top bar. (C) One representative UMAP for TCGA Glioma samples based on supervised model. Colors represent the unsupervised consensus classifications. (D) Scatter plots showing the percent of the 1000 models that cluster each sample (one dot) as belonging to each group. Samples classified as ambiguous by the supervised consensus are not shown here. (E) Kaplan–Meier survival estimate of TCGA overall survival data comparing IDHwt astrocytoma cases (dashed red), median survival = 19.9 months) to IDHmut-non-codel astrocytoma (green, median survival = 62.9 months) (LRT p = 2e-05). IDHwt glioblastoma (red) and IDHmut-codel oligodendroglioma (blue) are also shown as reference.