A key parameter for linkage-disequilibrium-based studies of human variation is the proportion of CNVs that can be tagged well by nearby SNPs. Such ‘taggability’ depends on CNV allele frequency and local SNP density, but not on CNV size (Supplementary Methods). Overall, the taggability of biallelic CNVs genotyped with high confidence seems to be largely similar to that of frequency-matched SNPs, except that rare CNVs are more poorly tagged; in CEU, 77% of CNVs >5% MAF are captured with r2 = 0.8, whereas only 23% of CNVs <5% MAF are similarly tagged. These results are similar to others in a smaller data set8. Interestingly, deletions are much better tagged by nearby SNPs than by duplications (average difference in maximum r2 is 0.25; P < 10−16), while controlling for allele frequency and local SNP density; this may be a result of the chromosomal dispersion of some duplications and an increased frequency of reversions and repeat mutations at some duplications36.
