Imputation of missing genotypes was based on HapMap Phase II genotypes for the European population (CEU).
The method of Gengler et al. (2007) for imputation of missing genotypes is based on the idea that the covariance between genotypes is proportional to the additive genetic relationship between individuals.
To infer genomic imprinting using such data sets, a multi-hierachical mixture model can be derived to estimate the missing parental genotypes based on observed offspring genotypes.
First, missing genotypes are sampled based on allele frequencies estimated from reference haplotypes, and then an Expectation-Maximization (EM) algorithm is used to estimate parameter values.
Because the majority of SNPs studied here are in low linkage disequilibrium, we were unable to accurately impute missing genotypes using methods based on linkage disequilibrium.
In SOLAR, each SNP genotype was converted to a covariate measure equal to 0, 1 or 2 copies of the minor allele (or, for missing genotypes, a weighted covariate based on imputation).
Class (5) probably had a high rate of true null alleles because we had data for multiple offspring and only included SNPs based on missing genotypes when at least five individuals had missing genotypes.
The incomplete SNP haplotypes of the component individuals were then completed by computational imputation of missing genotypes and of recombination breakpoints based on haplotype information from the nonmissing data.
To cope with missing genotypes, the computation of distances was based only on observed genotypes, that is: d ˜ i i ′ 2 = ∑ j ∈ S i i ′ (x i j − x i ′ j ) 2 / p i i ′ where p i i ′ represents the number of observed genotypes for lines i and i′.
A number of imputation software programs (fastPHASE [ 23], MACH [ 24], IMPUTE [ 25], Beagle [ 19], PLINK [ 26], DualPhase [ 27]) have been used to infer missing or untyped genotypes based on known information derived from flanking markers.
