Statistical imputation of untyped SNP genotypes based on reference haplotype panels can be used to overcome this challenge. Imputation has been primarily applied to increase the SNP density for analysis in studies where cases and controls were recruited together and genotyped in a uniform fashion on the same array at the same time, reducing the risk of batch effects that impact SNP genotype calling. In the case of composite public controls derived from multiple studies genotyped on different arrays, variations in genotyping protocols create systematic differences, which introduce the potential for differential error in estimated allele probabilities at each of the imputed markers and artifactual differences in allele frequencies. These artifacts might manifest in significant statistical bias in downstream tests of genotype–phenotype association.
