Multiple-breed genomic evaluation by principal component analysis in small size populations

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Genomic Prediction

Background:

• Accurate direct genomic values (DGV) are crucial for genetic selection in cattle.
• Understanding the impact of breed composition and predictor dimensionality is key for optimizing genomic prediction models.

Purpose of the Study:

• To investigate the effects of breed composition and predictor dimensionality on DGV accuracy in a multiple breed (MB) cattle population.
• To compare the principal component (PC) approach with the SNPBLUP model for calculating DGV.

Main Methods:

• Genotyped 3559 bulls from three breeds (Holstein, Brown Swiss, Simmental) using 54,001 SNPs.
• Calculated DGV using PC and SNPBLUP models across single breed (SB) and MB scenarios.
• Assessed prediction accuracy by correlating deregressed proofs with DGV in validation animals.

Main Results:

• In single breed scenarios, PC and SNPBLUP models yielded similar DGV accuracy.
• Multiple breed reference populations with the PC approach improved DGV accuracy for milk and protein yield in Brown Swiss cattle compared to SB-SNPBLUP.
• Generally, similar accuracies were observed between PC and SNPBLUP models when using MB reference populations, with random variation influencing results.

Conclusions:

• The principal component approach can enhance DGV accuracy in specific multi-breed contexts, particularly for certain traits.
• Breed composition and predictor dimensionality have nuanced effects on DGV accuracy, with potential for improvement in multi-breed scenarios.
• Further research is needed to fully elucidate the patterns and optimize genomic prediction across diverse cattle populations.