Extended Data Fig. 2: Defining intra-GSC transcriptional heterogeneity.

a, Heat map of Jaccard Index (more similar = blue, less similar = white) between marker gene lists across 86 intra-GSC clusters. A subset of 14 clusters, from 13 samples, display increased similarity (labelled as Cluster 1). b, Enriched pathways from 358 genes common to all 14 clusters defined in Extended Data Fig. 2a. c, Spearman correlation between inferred proportion G2M cells from scRNA-seq data vs. survival in an orthotopic xenograft model (left; n = 18 independent GSC xenograft models) and doubling time in vitro (n = 15 GSC cultures) in adherent (green) or neurosphere (orange) GSCs. Red line represents a linear regression line. Shaded grey area represents 95% confidence interval. d, 14 intra-GSC clusters share increased marker gene overlap and define a core proliferation module shared across 13 patients. Expression of select marker genes common across all clusters. Columns separated by intra-GSC cluster, bolded labels represent clusters with upregulation of the proliferation module. e, Relative expression of top 5 significant marker genes (based on logFC, one-sided Wilcoxon rank-sum test, FDR < 0.05) for clusters C1 and C2 within G549_L (left). UMAP visualization of select marker genes of C2 (right). f, Relative expression of top 5 significant marker genes (based on logFC, one-sided Wilcoxon rank-sum test, FDR < 0.05) for clusters C1-C5 within G837_L (left). UMAP visualization of select marker gene of C5 (right).