Genomic prediction based on data from three layer lines: a comparison between linear methods.

Area of Science:

• Animal genetics
• Quantitative genetics
• Genomic prediction

Background:

• Evaluating linear genomic prediction models for multi-line applications.
• Comparing genomic prediction to traditional pedigree-based BLUP.

Purpose of the Study:

• Assess the accuracy of various genomic prediction models in multi-line settings.
• Determine the impact of including multiple lines on prediction accuracy.

Main Methods:

• Compared BLUP, GBLUP, RRBLUP, RRPCA, BayesC, and Bayesian variable selection.
• Used laying hen data with phenotypes for egg production and known genotypes.
• Trained models on single or multiple lines (B1, B2, W1) with extensive SNP data.

Main Results:

• Genomic models outperformed pedigree-based BLUP by 0.13-0.16 in accuracy.
• Multi-line training marginally improved accuracy for B2, but not B1 or W1.
• RRPCA showed superior accuracy for B2; overall model differences were small.

Conclusions:

• Multi-line genomic prediction offers limited improvement for closely-related lines.
• Distantly-related lines may see reduced accuracy with multi-line prediction.
• RRPCA effectively addresses the "n < p" issue in genomic prediction.