The improved resolution of CNV breakpoints provided an opportunity to assess the extent to which distinct CNVs overlap in our data set. This is a complex problem in the absence of sequenced breakpoints for all variants, but we can use all validated CNVs, which may have some residual redundancy (that is, a single CNV could be split into two overlapping loci), to estimate an upper bound on this, and our non-redundant genotyped loci, which are probably biased against genotyping overlapping loci, to estimate a lower bound (Supplementary Methods). In this manner, we estimate the proportion of CNVs overlapping other CNVs to be in the range of 6% to 29%, which is far higher than the proportion of SNPs that are triallelic (that is, three different bases observed at the same site).
