Fig. 6: TAD subtypes.
a, t-distributed stochastic neighbor embedding result defining two major subtypes of TAD architecture, defined from the TAD edges-per-megabase according to each sample. b, Distribution of the average per-sample TAD size according to subtype. c, Example of TAD structure across all samples, with broad-subtype samples above the axis and narrow-subtype samples below it. Likewise, average contact frequency between locations for the same region, illustrating the reduced level of 3D interaction in broad-subtype samples. d–g, Distributions between TAD subtypes for the proportion of the genome defined by HMRs (n = 69) (d), 5hmC peak abundance (normalized to read depth, n = 52) (e), median TPM (n = 78) (f) and distributions between TAD subtypes for MYC CN (n = 76) (g). The center line represents the median; the box represents the upper and lower quartiles; and the whiskers represent 1.5× the IQR. P values were computed using a two-sided Wilcoxon rank-sum test. h, Significantly enriched hallmark pathways as determined by gene set enrichment analysis, showing an enrichment for MYC targets in the narrow-subtype samples. (i) Kaplan–Meier curves showing OS. NES, normalized enrichment score.
