Fig. 4

Cis-acting sequence features and trans-acting factors affect the degree of intron retention. (a) 5′ splice site scores, (b) 3′ splice site scores, (c) length, and (d) GC content of introns exhibiting increased (red) or decreased (blue) retention in cancer relative to normal controls. Each distribution illustrates the sample medians per cancer type. Dashed lines, median across all constitutive introns; dotted lines, median across 67,000 constitutive introns <1 kb (resulting in a length distribution similar to alternative introns). The median constitutive intron length of 1.4 kb is outside the range in (c). (e) Proportion of variation in intron retention that can be explained by a General Additive Model (GAM) of differential expression of RNA processing, transport and degradation genes. Each dot is a specific intron/mRNA combination. Gray box, combinations where more than one-third of the intron retention level can be explained by expression of trans-acting factors. (f) Overlap across cancer types for mRNAs within the gray box in (e), divided by Gene Ontology term. (g) Numbers of introns exhibiting increased or decreased retention in AML samples mutated in a specific gene compared to all wild-type AML samples for each gene. Dot size is proportional to mutation frequency. Analysis restricted to genes mutated in >5 samples. Mutations with >12 differentially spliced introns are highlighted. Note that the subgroup analyses illustrated here is statistically distinct from previous analyses, which were computed per-sample between matched tumor-normal pairs. This plot only illustrates introns that are consistently differentially spliced across the subgroup relative to other samples. (h) Numbers of introns exhibiting increased or decreased retention in breast cancer subtypes relative to all control breast samples (subgroup analysis, as in (g)). Analysis is over all 1,080 cancer and 104 normal breast samples. Dot size is proportional number of samples within each subtype