the same chromosome [56]. Equivalent arrays produced by Affymetrix had lower coverage of common SNPs for Europeans but higher coverage for Africans [55]. These figures are higher in denser arrays, with a clear gain of coverage in African subjects, so that for example the Illumina 610 provides approximately 60% coverage of the HapMap common SNPs with 80% correlation, meaning that the SNPs in the array correlate with 60% of HapMap common SNPs with r2 > 0.8. Recent studies have suggested that the gain of coverage using denser SNP arrays can almost be recovered by the imputation of untyped SNPs in Europeans and genotyping more samples with less dense arrays may increase the power of a GWAS [57]. Because imputation of untyped genotypes in subjects of African or Asian ancestry is less accurate [58], this strategy may not be useful in studies of non European subjects. These analyses used the HapMap catalogue of common SNPs as a gold standard. Estimates of the coverage of the full set of common and rare variants have been recently reviewed using sequence data and suggest that the initial calculations of the coverage of common platforms were overoptimistic [59]. Larger coverage of the real variants rather
