Genomic prediction accounting for dominance and epistatic genetic effects on litter size traits in Large White pigs

Abstract

Litter size traits of sows are crucial for the economic benefits of the pig industry. Three phenotypic traits of 1,206 Large White (LW) pigs, that is, the total number born (TNB), number born alive (NBA), and number of healthy piglets (NHP), were recorded. We evaluated a series of genomic best linear unbiased prediction (GBLUP) models that sequentially added additive effects (model A), dominance effects (model A+D), and epistatic effects (model A+D+AA, model A+D+AA+AD, and model A+D+AA+AD+DD) using chip data and imputed whole-genome sequencing (WGS) data to estimate genetic parameters and predictive accuracy. The reproductive traits of sows showed low heritability in this study, with narrow heritability of the three traits ranging from 0.030 to 0.064, and broad heritability ranging from 0.125 to 0.145. The inclusion of non-additive effects in the model improved the accuracy of genomic selection. In the chip data, compared with that of the A model, the A+D+AA+AD+DD model showed the greatest increase in accuracy for TNB, NBA, and NHP, with improvements of 1.78, 1.67, and 1.74%, respectively. Additionally, the accuracy of the imputed WGS data was greater compared to the chip data. For the TNB, NBA, and NHP traits, the predictive accuracy of the imputed WGS data improved by 3.26, 7.72, and 3.00%, respectively, compared with that of the chip data. In summary, these results suggest that non-additive effects in genomic selection could improve prediction accuracy and should be considered in pig genomic evaluation procedures.