Fig. 1: Global molecular differences in pediatric brain tumors by histologic type involving DNA methylation, gene expression, and structural variation.
a Hierarchical clustering of the top 2000 most variable CGI probes, as carried out on the CBTN DNA methylation dataset, involving 1744 tumors and 1536 patients. The differential patterns for the 2000 CGI probes and the top 2000 most variable genes are shown (each respective feature set involving different genes), with features respectively ordered by hierarchical clustering. As expected, the tumors are broadly segregated according to histologic type, although there is some variability in the grouping of tumors of a given histology26. See Methods regarding histology-based tumor type abbreviations. b Somatic SV breakpoint enrichment patterns by brain tumor histologic type. For each gene, the number of tumors with somatic SV breakpoints falling within 1 Mb of the gene start was tabulated, involving 1926 tumors and 1736 patients with WGS data. Represented are 2283 genes with significant breakpoint enrichment patterns for at least one of the histologic types represented (p < 0.0001, one-sided Fisher’s exact test). Represented along the bottom are any genes for which expression or CGI methylation is higher or lower with nearby somatic SV breakpoints (positive and negative correlation with SV breakpoints, respectively, p < 0.01 across all tumors by linear modeling of 1 Mb region by distance metric method18,48, correcting for both histologic type and gene copy number). Source data are provided as a Source Data file.
