Effect of breed composition in genomic prediction using crossbred pig reference population

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Genomic Prediction

Background:

• Conventional genomic prediction often focuses on single breeds, simplifying the process by avoiding population structure adjustments.
• Multi-breed genomic prediction, however, is susceptible to bias from population structure, necessitating appropriate modeling for accurate results.
• Crossbred reference populations present unique challenges for genomic prediction due to inherent population structure.

Purpose of the Study:

• To investigate the impact of population structure on multi-breed genomic prediction accuracy in crossbred pigs.
• To compare the effectiveness of different modeling strategies for population structure within genomic best linear unbiased prediction (GBLUP) models.
• To identify optimal approaches for enhancing genomic selection efficiency in commercial crossbred populations.

Main Methods:

• Evaluated five GBLUP models: no adjustment, principal component analysis (PCA) as fixed/random effect, and genomic breed composition (GBC) as fixed/random effect.
• Utilized crossbred pig datasets including Duroc × Korean native, Landrace × Korean native, and Landrace × Yorkshire × Duroc populations.
• Assessed prediction accuracy for backfat thickness and carcass weight under different population structure modeling scenarios.

Main Results:

• Models without population structure adjustment or with PCA/GBC as random effects (Models 1, 4, 5) achieved the highest prediction accuracies (e.g., backfat thickness: 0.59, carcass weight: 0.50).
• Models incorporating PCA or GBC as fixed effects (Models 2, 3) resulted in lower prediction accuracies (e.g., backfat thickness: 0.40-0.53, carcass weight: 0.34-0.38).
• The choice of population structure modeling significantly influenced prediction accuracy in multi-breed genomic prediction.

Conclusions:

• For multi-breed genomic prediction in crossbred populations, it is most effective to either not adjust for population structure or to model it as a random effect.
• Accurate accounting for population structure is critical for robust multi-breed genomic prediction frameworks.
• Optimizing prediction accuracy through appropriate modeling strategies can significantly enhance genomic selection efficiency and commercial production outcomes.