A gene-centric analysis of copy number variation reveals tumour suppressor modulation of cancer risk and novel associations with common diseases

Copy number variants (CNVs) are a significant source of genomic variation in humans. They can span large genomic regions that encompass whole genes and their flanking regions, having the potential to alter the number of functional copies in an individual. Typically, CNVs are analysed individually or combined into regions without regard to the genes they span. Here, we employ a gene-centric approach to link changes in whole-gene copy number to phenotype. We first examined the relationship between CNVs spanning whole genes and gene expression using data from GTEx and found that predicted effects on gene expression are often realised. Applying this method, we identified 34 gene deletions associated with cancer incidence, including several novel candidate associations. Most notably, we discovered that gene deletions in the 13q14 region are linked to increased cancer risk, a novel extension of a known somatic association to the germline, which we did not find using region-based methods. Statistical associations were found between whole-gene loss of tumour suppressor genes and overall cancer risk, as hypothesized. It was additionally found that the total number of whole-gene losses was strongly associated with increased cancer risk, even when known cancer-related genes are not included in the count. Finally, we extended these methods to non-cancer phenotypes, uncovering significant associations with common diseases, including asthma and heart attack.