The Application Of Single-step And Multiple-trait Genomic Selection In Chinese Simmental Cattle

Genomic selection means the marker assisted selection on a genome wide scale.Genomic selection can shorten the generation interval and improve the predictive accuracy of genomic estimated breeding values for Chinese beef cattle.Thus,the application of this strategy can shorten the gap of breeding with other countries by conducting genomic selection.This article focused on the problem of phenotypic missing and incomplete pedigree record in Chinese beef cattle breeding procedure.We conducted genomic selection based on the resource population of Chinese Simmental beef cattle,and finished the following aspects of innovative achievements.(1)This study was the first time to use REML(restricted maximum likelihood)to estimate the genetic parameters for 20 important economic traits including growth and development traits,carcass traits and meat quality traits,based on the genomic relationship constructed by the SNP information.These genetic parameters were the foundation of the next step studying on single step and multipletrait genomic selection.Our results revealed that daily gain and live weight of growth traits were medium and high heritabilities.Tenderloin,fore shank,biceps,outside,eye round,knuckle,inside cap off,hind shank and retail meat weight of carcass traits were medium and high heretabilities,and striploin,spencer roll and chuck roll were medium heretibilities.Back fat thickness of carcass traits was low heritability.Eye muscle area of meat quality traits was high heritability and shear force was low heritability.Totally,the heretabilities of 20 traits were between 0.15 and 0.62.(2)We conducted single step genomic selection based on simulation dataset and Chinese Simmental beef cattle,and investigated the impact of missing phenotypes on genomic predictive accuracies.We also looked for the best weighted parameter of the adjusted single step model for the important economic traits in Chinese Simmental beef cattle.Simulation study revealed that when the population was 1000 and the phenotype was miss 5% percentage,the predictive accuracy reduced 0.3~3.3 percentage points.And the predictive accuracies reduced 0.7~4.5 percentage points when the phenotype was missing 10%.Moreover,the predictive accuracy for low heritability trait with the phenotype missing declined more than these for high heritability.Using real dataset,the predictive accuracies of 20 traits in Chinese Simmental beef cattle for GBLUP were from 0.10 to 0.46,and the predictive accuracies for ss GBLUP were from 0.13 to 0.49.For most traits,ss GBLUP achieved higher predictive accuracies than GBLUP.When the weighted parameter of the adjusted single step model was 0.4,the average predictive accuracies of six important economic traits including daily gain weight,live weight,carcass weight,retail meat weight,dressing percentage and lean meat percentage trait was the highest,therefore,this parameter could be considered as the best value in Chinese Simmental beef cattle population.(3)We conducted the multiple-trait genomic selection in Chinese Simmental beef cattle population and investigated the impact of phenotype missing on predictive accuracy.When there was no phenotype missing,the predictive accuracies of multiple-trait Bayes BLUP and Bayes A model were much higher than those for single-trait genomic selection,especially for the low heritability traits.The average improvement accuracy was 1.4 percentage points and 0.8 percentage point for mt Bayesblup and mt Bayes A.And the highest improvement accuracy was 4.8 percentage points for shear force trait.The predictive accuracy for the high genetic correlation traits also improved a lot,for example,the high genetic correlation traits between liveweight and carcass weight.For other traits,multipletrait genomic selection model produced similar predictive accuracy with single-trait genomic selection model.However,when there were some phenotypes missing for the 13 important economic traits,multiple-traits genomic selection model achieved 0.2~5.4 percentage points higher accuracies than single-trait model.And the predictive accuracies improved more for the low heritability traits than high heritability traits.