Fig. 2: Transcriptional regulation of malignant cell-type transitions.
a, t-SNE visualization of n = 2,520 malignant ETMR cells colored for cells predicted as basal progenitor cells (green) using a random forest classifier trained on cells from the fetal neocortex. b, t-SNE visualization of n = 2,520 malignant ETMR cells, highlighting normalized expression of the basal progenitor cell marker genes ASCL1, INSM1, NHLH1 and NEUROD1. c, Heatmap showing normalized area under the curve (AUC) values (indicating regulon activity) from the SCENIC analysis. Four groups of regulons were formed by hierarchical clustering across patients. The annotation above the heatmap indicates the classification of n = 2,520 cells into three malignant cell types. d, Network showing similarity scores (thickness of connecting lines) determined using pairwise comparison of regulons shown in c. e, t-SNE visualization of n = 2,520 malignant ETMR cells colored according to refined cell types. f, t-SNE visualization of n = 2,520 malignant ETMR cells colored according to cell cycle state. g, t-SNE visualization of n = 2,520 malignant ETMR cells overlaid with differentiation trajectories generated by RNA velocity analysis of a representative ETMR sample (BCH736). h, Representative multiplexed IF images of n = 3 different primary ETMR samples for the marker TFs PAX6 (NSC-like cells, red), NEUROD1 (intermediate cells, green) and PEG3 (neuron-like cells, magenta; top). Magnification of the indicated region (dashed outline) in ETMR3 is shown as pairwise contrasts (bottom). i, Pseudo image (left) showing a reconstruction of cells analyzed using in silico phenotyping for the same region as in h, colored according to marker gene expression. The violin plots show the nearest neighbor distance analysis between different ETMR cell types identified using multiplexed phenotyping. A total of n = 46,621 cells were phenotyped. P values were calculated using a two-sided, unpaired Student’s t-test. The lines represent the mean values. h, Scale bars, 300 µm.
