Investigating the impact of multiallelic copy number variation on disease-relevant traits in UK and African populations

The contribution of multiallelic CNVs (mCNVs) to disease risk has not been widely studied, as they have been difficult to characterise at large-scale genome-wide, and are often not strongly associated with flanking SNPs. Previous work in this area frequently comes to conflicting conclusions. Improved understanding of the role of mCNVs in disease risk could lead to novel insights into the pathobiology of disease.I used ClinCNV to robustly type 69 mCNVs from UK Biobank whole exome sequences in discovery (n=152,853) and replication sets (n=269,317). Discovery and replication PheWAS of these variants used the DeepPheWAS R package, which interrogates clinically-curated composite phenotypes by integrating self-report, primary and secondary health care data. Unrelated participants of African (AFR), European (EUR) and Central/South Asian (CSA) ancestry were included in this analysis.In total, 173 significant mCNV-phenotype associations were detected from 26 mCNVs (FDR < 0.01 for EUR, FDR < 0.05 for AFR and CSA), of which 114 associations replicated (p < 0.05). Four of eight potentially novel mCNV-phenotype signals were independent of neighbouring associated SNPs. mCNV breakpoints were estimated in whole genome and long read data; validating copy number genotypes and confirming whether the variant partially or entirely overlapped the gene, to possible functional effects.Statistically significant correlation of mCNV copy number with gene expression or protein level was detected in 33% of gene expression corelations and 80% of protein correlations. While predicted copy number generally had a weak effect on gene expression and protein levels (correlation < 0.3), 25% of gene expression and 30% of protein level correlations were moderate to strong. These findings support the biological relevance of mCNVs, which may underlie phenotype mechanisms. Accurate typing and large sample sizes are necessary to reveal genetic associations of mCNVs, and this work provides several novel likely causative variants contributing to clinically-important traits.