We explored whether the CNVs from this study might be plausible candidates for causal variants for known complex trait associations from genome-wide association studies (GWAS). We examined 1,554 trait-associated SNPs from 279 publications (NHGRI GWAS website47, downloaded on 15 June 2009), In the CEU, 474 out of 1,521 polymorphic trait-associated SNPs fell within a recombination hotspot interval that also contained a CNV. We then examined whether the CNVs in these intervals were in strong linkage disequilibrium with the trait-associated SNP in the different HapMap populations. For genotyped biallelic CNVs we assessed linkage disequilibrium using correlation (r2) within phased haplotypes, but to include multiallelic and ungenotyped CNVs in this analysis we also considered the squared Pearson correlation coefficient between the SNP genotypes and the copy number intensity data. We identified 34 trait-associated SNP to CNV correlations with an r2 of greater than 0.5, at 30 loci across 22 traits (Fig. 5c, Table 2 and Supplementary Fig. 1.12), five of which were found in the HLA. These CNVs include three previously identified CNV-trait associations13–15, which represent all the positive controls for this
