Fig. 2

Three MES states of GBM. A Boxplots show the Pearson correlations of the scores of each pair of GBM function-specific signatures across MES cells, where each dot represents the correlation in one patient. Significance defined by t-test is indicated with asterisks (∗∗∗∗, p < 0.0001; ∗∗∗, p < 0.001; ∗∗, p < 0.01; *, p < 0.05). In each boxplot, horizontal line shows the median, the box borders show upper and lower quartiles, the vertical line indicates the 5th through 95th percentiles. B Heatmap depicts the Pearson correlation between pairs of MES programs, averaged across patients. C Dot plot shows the PC2 and PC3 scores of MES cells from all of the samples from the Neftel et al. dataset, colored by the difference between their score for MES-Hyp and MES-Ast. D Top: Heatmaps of 4 samples from Neftel et al. X-axis represents MES cells ordered by their PC3 score and Y-axis shows highest and lowest scoring PC3 genes. Selected genes from MES-Ast (top) or MES-Hyp (bottom) are labeled. Bottom: Heatmap shows the average score of each cell for the core MES, MES-Hyp, and MES-Ast programs. E Main heatmaps show the score of the MES cells from each tumor for the different MES programs. The top heatmap shows the average (uncentered) score and the bottom heatmap shows the score after centering each row (across samples). Annotation bars show the glioma type (top) or the sample’s TCGA-subtype (bottom). F Distributions of the difference in MES-Hyp and MES-Ast scores for individual tumors, colored by glioma type as in E). G Bars show the correlation between the expression of NF1 and the core MES scores (orange), and the partial correlation of NF1 expression with core MES scores after controlling for MES-Hyp scores (yellow) or for MES-Ast scores (brown), across TCGA bulk RNA-seq samples. Error bars correspond to first and third quartiles based on sampling 75 tumors with 1000 iterations; the difference between averages is significant by Wilcoxon rank sum test (p=2.2e−16) as indicated by asterisks