Optimization Study Of Genomic Pre-selection On Swine

Genomic selection(GS) is hardly applied to real swine breeding industry currently. But there are many opportunities for GS to achieve the goal. One of them is called genomic pre-selection, which pre-selects small swine population to later stage of performance measuring based on genomic estimated breeding value (GEBV), thus saving costs of genetic estimation. Three different genomic selection strategies were designed in this thesis. Strategy one was totally based on directly genomic selection, which selected next breeding population according to GEBV. The GEBV was calculated based on individuals’genotypes; Strategy two pre-selected a part of individuals to later stage of performance measuring based on GEBV. After the performance measurement information coming out, selecting was proceeding via conventional BLUP; Strategy three was improved based on strategy two. In strategy three, candidates owned phenotypes and GEBV information after performance measurement. In this case, a bivariate evaluation model was designed to composite these information to increase the accuracy of selection. Three strategies were tested and verified in two simulated traits datasets of swine (days to100kg and back-fat to100kg). In addition, five GEBV estimating methods (GBLUP, Bayes A, Bayes B, Bayes C and Bayes C π) were compared in the aspects of selecting effect of the swine datasets. Several levels of key factors(proportion of pre-selecting, dense of markers and heritability) in genomic pre-selection were optimized. Results are as follows:(1)For three strategies, genomic pre-selecting strategies (strategy two and three) using phenotype information obtained higher accuracy than the one (strategy one) directly based on GEBV. The modified two-traits estimating model acted out the best of accuracy and genetic gain under the basic parameter set;(2)For five genomic breeding evaluating methods, they turned out similar accuracy of selecting and genetic gain in strategy two and three. GBLUP got the lowest accuracy of selecting and Bayes A was slightly superior to it. Bayes B, Bayes C and Bayes Cπ got approximate value and better than previous two methods; Bayes Cπ kept optimum stability; (3)As for key factors affecting pre-selection, when boar pre-selected30%showed steady high accuracy, better than10%and50%; when sow pre-selected50%and70%resulted comparative accuracy,much better than100%; Genetic gain had the same changes corresponding to accuracy. Accuracy of60k markers transcended3k markers. Each method of selection accuracy rose along with the higher heritability.It is concluded that, considering GEBV and phenotype information at the same time can increase the selecting accuracy, thus letting breeding population obtain more genetic gain. Bayes B, Bayes C and Bayes Cπ show high accuracy of selecting and steady selecting effect in swine datasets. Pre-selection before performance measurement may influence the accuracy of the whole process, especially for boar, appropriate selecting proportion increase accuracy. In this thesis, the simulated data were based on real swine population datasets, optimal strategy was presented, preferably method of estimating GEBV was tested and verified, and relatively appropriate pre-selection proportion was confirmed. All of these would be guiding significance in practical application of swine genomic selection.