Fig. 3

Genes with altered expression associated with nearby SSV breakpoint according to cancer type. a For each cancer type, numbers of significant genes showing correlation between expression and nearby SSV breakpoint (FDR < 10% by distance metric method, linear model correcting for CNA). b The X-axis indicates the FDR in the most significant of the 23 cancer types. The Y-axis indicates the FDR when the 2334 cases are analyzed as a combined pan-cancer cohort. Genes in the upper left quadrant reached significance only in the pan-cancer analysis. Genes in the lower right quadrant reached significance only in one or more single-type analyses. Genes in the upper right quadrant were significant in both the pan-cancer set and in individual cancer types. The color of data points represents the most significant cancer type (following a color scheme). c For each cancer type, numbers of DNA methylation probes (Illumina 450K array platform) targeting CpG islands (CGIs) with high methylation in the given cancer type versus other cancer types (FDR < 0.001, t test using logit-transformed data), for which the associated gene also shows a positive correlation between expression and nearby SSV breakpoint for that same cancer type (FDR < 0.1, from a). The numbers of CGI probes expected to overlap by chance between the differential methylation results and the expression vs SSV results are also indicated (gray bars), along with any significance of overlap represented by the actual results (asterisks, p values by chi-squared test). d For 893 CGI DNA methylation probes showing both high cancer type-specific methylation and significant positive correlation between expression and SSV breakpoint for any one of the 20 cancer types surveyed (from c), the associated SSV versus expression correlations (from a), average DNA methylation by cancer type, and differential methylation in each cancer type versus other cases (t-statistic using logit-transformed values). See also Additional file 1: Figure S4 and Additional file 6