Abstract: The aims of current study were to show the process of genomic imputation and investigate the factors affecting accuracy of genomic imputation.The data of 50K panel imputed to 150K for dairy cattle was used to compare accuracy and concordance of imputation for two software (Beagle 5.1 and Minimac 3).Concordance was calculated by cross validation of individuals with imputed data and real data.The target population for imputation should be phased by Minimac 3 and the function of Beagle 5.1 including pre-phasing and imputation.The correlation of concordance between Minimac 3 and Beagle 5.1 were 0.98.For Beagle 5.1,the average of imputation accuracy (r²) and concordance were 0.9841 and 0.9914,respectively,and the correlation between imputation accuracy and concordance was 0.39.For Minimac 3,the average of imputation accuracy (r²) and concordance were 0.9782 and 0.9911,respectively,and the correlation between imputation accuracy (r²) and concordance was 0.36.Imputation accuracy (r²) was associated with minor allele frequency due to the formula for calculating accuracy from the software.With the increasing of minor allele frequency and heterozygosity for makers,the concordance of imputation was decreased.There was a steep decline when heterozygosity was higher than 0.6 (concordance of imputation was lower than 0.8).However, the accuracy of Beagle 5.1 software was better than that of Minimac 3 software under the same minor allele frequency and heterozygosity of imputed site. The accuracy of imputation (r²) was mainly affected by heterozygosity of SNPs and Beagle 5.1 had better performance on imputation accuracy than that of Minimac 3. Using concordance as the accuracy of imputation to select SNPs could avoid losing useful makers for further study.

Key words: Holstein cow; population; genomic imputation; accuracy