Effect of Lower Density SNP Panel Size on Accuracy of Imputation and Genomic Estimated Breeding Values in Pigs

Application of genomics has been limited in swine industry due to cost of the high-density (HD) SNP panels. Genotyping using low-density (LD) panel and imputation to a HD panel can be a viable solution. The objectives of this study were to investigate the effect of panel density on imputation accuracy and precision of the predicted genomic estimated breeding values (GEBVs). Estimated breeding values and Illumina Porcine60K BeadChip genotypes were available for 1,378 Duroc (DU), 2,361 Landrace (LA), and 3,192 Yorkshire (YO) pigs. A total number of 931, 1,631, and 2,103 pigs were used as reference population for the DU, LA, and YO, respectively, and the remainders used as validation population for each breed respectively. An LD panel of 3,000 evenly spaced SNPs had high imputation concordance rates of 93.78% for DU, 97.07% for LA, and 97% for YO, and yielded very high correlation (>97%) of the predicted GEBVs between the truly observed 60K SNPs and the imputed data. The imputation accuracy was influenced by the reference population size and parental genotype information in the reference population. The parental genotype information became less crucial when the LD panel had 3,000 SNPs or more. Overall, the imputation accuracy rates increased substantially from the 300 to 3,000 SNP LD panel and became steady when panel size increased to more than 3,000 SNPs. The correlation of the GEBVs generated using true and imputed SNPs generally increased as the imputation accuracy rates were improving.