Fig. 3

Intron retention preferentially affects genes encoding RNA processing factors. (a) RNA-seq read coverage of FUS for patient-matched tumor and normal samples from breast and colon. Shaded boxes indicate differentially retained introns. (b) Histograms illustrating the numbers of introns exhibiting increased (red) or decreased (blue) retention in cancer relative to normal samples, computed across all cancer types. (c) Hierarchical clustering of all retained introns (rows) and all cancer types (columns). Analysis restricted to introns that exhibit increased (red) or decreased (blue) retention relative to normal controls in >10 % of samples for at least one cancer type. Clustering is based on Euclidean distances computed over intron retention frequencies and Ward’s agglomeration method. (d) The combined -log₁₀ false discovery rate of the most significant Biological Process Gene Ontology (GO) terms enriched among genes containing differentially retained introns in at least 20 % of samples within each cancer type. Colors as in Fig. 1. (e) Percentage of samples within cancer types with differential intron retention for select genes mapped to the ‘mRNA export from nucleus’ GO term (GO: 0006406). Dashed line, average across all genes and cancer types. (f) Distribution of Pearson correlation coefficients between intron retention and gene expression across all samples within each cancer type. Dashed line, median taken over all samples for each cancer type. Colors as in Fig. 1. (g) Scatter plots comparing intron retention to fold-change of the corresponding parent genes for two colon adenocarcinoma samples relative to their patient-matched normal control