We performed genotype imputation in the MRCA and MRCE samples independently. We used the ILMN300K genotypes to mimic the data that might be used in a typical GWAS and to impute the polymorphic SNPs in the Phase II HapMap. We imputed genotypes for all polymorphic HapMap SNPs by using a hidden Markov model programmed in MaCH (Li et al. 2010). The method combines genotypes from the study samples with the HapMap CEU sample (July 2006 phased haplotype release) and identifies the stretches of haplotype shared between the study samples and the HapMap sample. For each individual, the genotype at the untyped SNP can be summarized by taking (1) the most likely genotype according to the posterior probability of the three possible genotypes and (2) allele dosage, the expected number of copies of the reference allele (a fractional value between 0 and 2). To take into account the uncertainty of imputation, we used allele dosage in downstream analysis.
