A Comparative Study of Optimizing Genomic Prediction Accuracy in Commercial Pigs

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Livestock Breeding

Background:

• Genomic prediction (GP) is vital for accelerating genetic gain in livestock and plant breeding.
• GP accuracy is influenced by statistical models, marker density, and cross-validation methods.
• Optimizing GP is crucial for economically important traits in commercial pig populations.

Purpose of the Study:

• To evaluate and optimize genomic prediction accuracy for eight key carcass and body traits in a Duroc × (Landrace × Yorkshire) pig population.
• To compare the performance of various statistical models, including GBLUP, ssGBLUP, and Bayesian approaches.
• To assess the impact of marker density and cross-validation strategies on prediction accuracy.

Main Methods:

• Utilized 50K SNP chip data imputed to whole genome sequence (WGS) level from 1494 DLY pigs.
• Compared seven distinct genomic prediction models: GBLUP, ssGBLUP, and five Bayesian models.
• Employed cross-validation strategies with varying numbers of folds and analyzed different marker densities.

Main Results:

• The ssGBLUP model demonstrated superior performance across all evaluated traits, achieving prediction accuracies between 0.371 and 0.502.
• Prediction accuracy increased with higher cross-validation folds and marker density, especially in low-density panels.
• Diminishing returns were observed for prediction accuracy with medium-to-high-density marker panels.

Conclusions:

• The ssGBLUP model is highly effective for genomic prediction of carcass and body traits in DLY pigs.
• Strategic selection of statistical models, marker density, and cross-validation is essential for maximizing GP accuracy.
• Findings provide valuable guidance for enhancing genetic progress in commercial pig breeding programs.