Fig. 6

Global alterations in transcription and DNA methylation associated with the overall burden of structural variation across cancers. a Numbers of significant genes (FDR < 5%), showing correlation between expression and the total number of SSV events detected across the 2334 cases with RNA-seq data. Linear regression models evaluated significant associations when correcting for specific covariates (in addition to low-pass versus high-pass WGS), as indicated. b Numbers of significant CGI probes (FDR < 5%), showing correlation between DNA methylation and the total number of SSV events detected across the 1482 cases with methylation data. Linear regression models evaluated significant associations when correcting for specific covariates (in addition to low-pass versus high-pass WGS), as indicated. c Scatter plot of global SSV index (measuring total number of SSV events, correcting for high-pass versus low-pass WGS) versus overall methylation (median beta of all 450K probes within the sample profile). p value by linear model correcting for cancer type. d Significantly enriched GO terms for genes correlated (FDR < 1%, with corrections for cancer type, CNA, and low-pass versus high-pass WGS) with the total number of SSV events. p values by one-sided Fisher’s exact test. e Across the 1482 cases with DNA methylation data, with cases ranked high to low by global SSV index quartiles, selected molecular features are represented, including top expression correlates with total number of SSV events (from d), CGI probes with DNA methylation high with total number of SSV events (FDR < 1%, correcting for cancer type, low-pass versus high-pass WGS, CNA, proximal BP pattern, age, and overall methylation) and with associated mRNAs low with total number of SSV events (as well as anti-correlation between mRNA and methylation, Pearson’s FDR < 10%), overall methylation (from c), tumor purity, tumor ploidy, aneuploidy [29], overall CNA, exome mutation rate, and patient age. Expression and methylation values are normalized or centered within each cancer type. Highlighted genes are represented by multiple CGI probes. p values by linear model correcting for both cancer type and low-pass versus high-pass WGS. See also Additional file 1: Figure S8 and Additional file 11