Studies On The Efficiency Of Genomic Selection By Stochastic Process

The main purpose of this research is to study the efficiency of Genomic Selection method by computer simulation programming. And we compare the estimating accuracy and genetic gain brought by different individual breeding value estimating methods. we selected and maintained animals according to their estimated breeding value by different methods for 20 generations.Two kinds of selection intensity, the low intensity has the fraction selected of male being 5%, and that of female being 20%. We simulated 20 selected generations after the form of ideal population, and calculated the rate of inbreeding and genetic gain for each selection method. This programming experiment set three levels of heritability which was 0.3, 0.5 and 0.8. We compared the rate of inbreeding and estimating accuracy of 6 methods, which included the phenotypic selection, ICE(Iterative Conditional Expectation) method ,ECM(Expectation Conditional Maximization)method , LS(Least Squares)method and GS-BLUP(Genomic Selection Best Linear Unbiased Prediction).Results are:(1) Estimation accuracy of EBV: the order of estimation accuracy is: ECM2 > ECM1 and ICE > GS-BLUP > phenotypic selction > LS. ECM2 is 0.72～0.85, ECM1 and ICE are 0.68～0.79, GS-BLUP is 0.61～0.79,phenotypic selection is 0.30～0.78,LS is 0.29～0.57. The accuracy increases with the increase of heritability. Different selection intensity and updating the SNP effects have no effect on the accuracy. As the selection progressing, the estimation accuracy decreases.(2)Inbreeding coefficient: As the selection progressing, the inbreeding coefficient increases. It also increases with the increasing of selection intensity. When not updating SNP effects, different selection intensity and heritability have no effect on the coefficient. In the low selection intensity, the coefficient can reach 0.16 at the generation 1020; in the high selection intensity, it can reach 0.50. When updating SNP effects, the order of inbreeding coefficient is: ECM2 > GS-BLUP > LS > phenotypic selection. In the low selection intensity, the coefficient can reach 0.20 at the generation 1020; in the high selection intensity, it can reach 0.55. (3) Inbreeding rate: As the selection progressing, the inbreeding rate increases. The higher the selection intensity, the larger the inbreeding rate. When not updating SNP effects, different selection method and heritability have no effect on the inbreeding rate. In the low selection intensity, the rate is 0.006～0.013. In the high selection intensity, the rate is 0.03～0.05. It is larger when updating SNP effects than not updating SNP effects. The order of inbreeding rate is: ECM2 > GS-BLUP > LS > phenotypic selection. The higher the heritability, the smaller the difference among different selection method. In the low selection intensity, the rate is 0.007～0.015. In the high selection intensity, the rate is 0.03～0.06.(4) Genetic gain: When not updating SNP effects and low selection intensity, the accumulated genetic gain of phenotypic selection can reach 6.5～6.8 in generation 1020, GS-BLUP 4～5, LS 2～3,other methods around 4. In the high selection intensity, The genetic gain of phenotypic selection can reach 5.5～7 in generation 1020, GS-BLUP 4～5.5, LS 2～4,other methods around 4. When updating SNP effects, each method has the same accumulated genetic gain, 5～6.8 in the low selection intensity, 5～7 in the high intensity.In conclusion, ECM2 is recommended for the short term breeding scheme, and when it comes to longer term, GS-BLUP is recommended.