Abstract
Common genetic variation throughout the genome and rare coding variants identified to date explain about half of the inherited genetic component of epithelial ovarian cancer risk. It is likely that rare variation in the noncoding genome will explain some of the unexplained heritability, but identifying such variants is challenging. The primary problem is a lack of statistical power to identify individual risk variants by association, as power is a function of sample size, effect size, and allele frequency. Power can be increased by using burden tests, which test for the association of carriers of any variant in a specified genomic region. This has the effect of increasing the putative effect allele frequency. PAX8 is a transcription factor that plays a critical role in tumor progression, migration, and invasion. Furthermore, regulatory elements proximal to target genes of PAX8 are enriched for common ovarian cancer risk variants. We hypothesized that rare variation in PAX8 binding sites is also associated with ovarian cancer risk but unlikely to be associated with risk of breast, colorectal, or endometrial cancer. We have used publicly available, whole-genome sequencing data from the UK 100,000 Genomes Project to evaluate the burden of rare variation in PAX8 binding sites across the genome. Data were available for 522 ovarian cancers, 2984 breast cancers, 2696 colorectal cancers, 836 endometrial cancers, and 2253 noncancer controls. Active binding sites were defined using data from multiple PAX8 and H3K27 chromatin immunoprecipitation sequencing experiments. We found no association between the burden of rare variation in PAX8 binding sites (defined in several ways) and risk of ovarian, breast, or endometrial cancer. An apparent association with colorectal cancer was likely to be a technical artifact as a similar association was also detected for rare variation in random regions of the genome. Despite the null result, this study provides a proof-of-principle for using burden testing to identify rare, noncoding germline genetic variation associated with disease. Larger sample sizes available from large-scale sequencing projects, together with improved understanding of the function of the noncoding genome, will increase the potential of similar studies in the future.