Imputation of medium density genotypes from custom low density genotype panel in sheep

Abstract Text: A custom low density genotype panel has been developed with 16,351 single nucleotide polymorphisms (SNPs), 12,118 of which are on the medium density Illumina Ovine50 Beadchip which has 51,135 SNPs. The objective of the present study was to quantify the accuracy of imputation from the low density to medium density panel in five different sheep breeds. Medium density genotypes were available on 2375 sheep from the breeds Suffolk (n=566), Texel (n=318), Vendeen (n=461), Charollais (n=559) and Belclare (n=471). The youngest 75 animals per breed were used as the validation population; the 37,278 autosomal SNPs on the medium density panel that are not on the low density panel were masked for the validation population. Imputation was undertaken across the entire genome simultaneously using both family and population wide linkage (disequilibrium) information. Concordance rates and the correlation between the true and imputed genotypes were estimated for the validation animals which included the low density SNPs in the calculation. Across all genotypes, the correlation between the actual and imputed genotype was 0.983; the mean genotype (allele) concordance rate was 0.979 (0.989). The mean genotype and allele concordance rate per individual varied from 0.864 to 0.997 and from 0.929 to 0.999, respectively. The individual with the poor concordance rate was an outlier and the minimum genotype (allele) concordance rate excluding this individual was 0.920 (0.958). Mean genotype concordance rate per breed was 0.984, 0.972, 0.982, 0.969 and 0.989 for Belclare, Charollais, Suffolk, Texel and Vendeen, respectively. Imputation accuracy not accounting for pedigree was marginally better than when pedigree was accounted for in the imputation process. Imputation accuracy with a reference population of only the breed of animal to be imputed was also marginally better than when multiple breeds were included in the reference population; imputation accuracy of breeds not represented in the reference population were considerably worse. The low density panel is therefore a useful, lower cost, strategy to achieve genomics evaluations in these sheep breeds.

Keywords: imputation, sheep, genomic, low density