Study On Genomic Single-step Method With Metafounders In Danbred Pig Populations

Single-step genomic best linear unbiased prediction(SSGBLUP),as known as single-step genomic selection method,is the standard genomic evaluation method used in breeding companies around the world,which plays an important role in genetic improvement in production of animals.Issues like pedigree and genomic relationship matrices not compatible directly exist in this method,which decrease the effect of its application.To save the problems,the theory of metafounder was proposed as a new thought of the development of SSGBLUP by Lagarra et al.Researches of Single-step genomic best linear unbiased prediction with metafounders(MF-SSGBLUP)are few in Average daily gain(ADG)and Feed conversion ratio(FCR)at present.In addition,the best weight ratio of pedigree and genomic relationship matrices is unknown in SSGBLUP and MF-SSGBLUP.Therefore,the aim of the study is to compare the genetic parameter,predictive ability and bias of genomic evaluations obtained with MF-SSGBLUP and standard SSGBLUP using univariate and bivariate models for ADG and FCR in Dan Bred Landrace(LL)and Yorkshire(YY)populations.It also aimed to explore the effect of different weight factors ω of ADG and FCR,and figure out the optimal weight factors for optimizing genomic evaluation process and giving guidance for production practice.The results of the study are shown below.(1)Genetic parameters estimated by SSGBLUP and MF-SSGBLUP methods with univariate and bivariate models were compared for ADG and FCR in LL and YY pigs.In LL population,for the same trait,the estimates of the additive genetic variances were similar for the four methods.Heritabilities were nearly the same.The genetic correlation between ADG and FCR using a bivariate model with MF-SSGBLUP was estimated higher negative than that with SSGBLUP method.In YY pigs,estimates of the additive genetic variances for ADG and FCR were higher with MF-SSGBLUP than with SSGBLUP methods.The genetic correlations between ADG and FCR with two bivariate models were both estimated as-0.46.(2)Using different ω ranging from 0.05 to 0.95 with an interval of 0.05,cross-validation method was used in genomic evaluation for ADG and FCR in LL and YY pigs.For ADG,The predictive accuracies of the four genomic evaluation methods were similar in the same population.The predictive accuracies of the univariate models(0.355)were slightly higher than that of bivariate models(0.344)for FCR in YY population.With the optimal ω,the predictive accuracies of the four methods were similar of a subgroup,and genotyped subgroups performed better than non-genotyped subgroups.(3)Genomic predictive biases were compared using different ω in four genomic evaluation methods for ADG and FCR in LL and YY pigs.Overall,The regression coefficients increased with the increase of ω.In LL population,the regression coefficients of ADG were nearly the same for the four methods.The unbiasedness for FCR estimated by bivariate models were better than that estimated by univariate models.In YY population,for ADG,MF-SSGBLUP method with bivariate models showed the optimal unbiasedness among the four methods.The unbiasedness of genomic prediction were better for non-genotyped subgroups than genotyped subgroups for the four methods with the optimal weight factors ω.In our study,for ADG and FCR in LL and YY pigs,MF-SSGBLUP methods with bivariate models showed the best effect of genomic evaluation,and the optimal weight factors were slightly different between SSGBLUP and MF-SSGBLUP methods.