Fig. 4

CpG islands (CGIs) with altered DNA methylation associated with nearby SSV breakpoint. a For each of the indicated genomic region windows in relation to genes associated with CGIs, numbers of significant CGIs (FDR < 5%, correcting for both cancer type and gene-level CNA), showing correlation between DNA methylation and associated SSV event, across 1482 cases with WGS and DNA methylation data. Results from distance metric method (± 1 Mb) as well as genomic region windows methods are shown. CGI DNA methylation probes tested for the given region had at least three cases with SSV breakpoint. b Fraction of promoter-associated CGIs, for the CGIs associated with increased methylation (by distance metric method from a), and for the CGIs associated with decreased methylation. p values by chi-square test. c Breakdown by probe position relative to gene, for the CGIs associated with increased or decreased methylation, respectively. p values by chi-square test. d Overlap between CGI probes with SSV-associated altered methylation and nearby genes with corresponding SSV-associated altered expression. For expression- and DNA methylation-SSV breakpoint associations inverse to each other for the same genes (e.g., gene expression positively correlated and DNA methylation negatively correlated with nearby breakpoint, using FDR < 10%), the subset of CGI methylation probes for which the associated genes are negatively correlated between methylation and expression (FDR < 10% by Pearson’s correlation on logit- or log-transformed values) across the 1482 cases are indicated. Genes highlighted are represented by multiple CGI probes. p value for significance of overlap by chi-squared test. e As an example of a significant gene, gene expression levels (left) and DNA methylation levels (right) of FASN, corresponding to SSVs located in the genomic region 1 Mb downstream to 1 Mb upstream of the gene. Each point represents a single case (closest SSV breakpoint represented for each case). Cases with gene amplification are indicated. f A set of 143 CGI methylation probes significantly associated with nearby SSV breakpoints for one or more individual cancer types (FDR < 10%), for which the expression versus SSV association for the corresponding gene within the same cancer type are significant (FDR < 10%) and in the opposite direction. Left panel represents SSV-methylation associations; right panel represents SSV-expression associations. FDR values by distance metric method, correcting for CNA. Gene name coloring corresponds to associated cancer type. See also Additional file 1: Figures S5 and S6 and Additional file 7