One use of imputation is to combine data for genome-wide association studies performed using different array platforms. Therefore, we first measured the change in performance of imputation for common (array-based) SNPs using a HapMap 3 panel of 410 phased European-ancestry chromosomes (CEU+TSI) in comparison with a HapMap Phase II panel of 120 CEU chromosomes (HMII-CEU). Each panel was used to impute array SNPs in 1,393 Europeans of the 1958 British birth cohort (58BBC), which had previously been genotyped using earlier versions of the Affymetrix and Illumina chips16,17. Using the Illumina array genotypes, we imputed HapMap 3 SNPs on chromosome 20 and calculated the mean r2 between true (called) genotype and imputed genotype dosage for each Affymetrix SNP not on the Illumina chip (Supplementary Table 8).
