Different genomic relationship matrices for single-step analysis using phenotypic, pedigree and genomic information

Area of Science:

• Animal Breeding and Genetics
• Genomic Selection
• Quantitative Genetics

Background:

• Single-step genomic evaluation integrates pedigree and genomic data.
• Compatibility between genomic and pedigree relationship matrices is essential.
• Optimal construction of genomic relationship matrices remains an area of investigation.

Purpose of the Study:

• To evaluate different methods for constructing genomic relationship matrices.
• To assess the impact of scaling genomic relationships on breeding value estimation.
• To determine the optimal approach for single-step genomic evaluations.

Main Methods:

• Analyzed litter size data from 338,346 sows.
• Constructed genomic relationship matrices using various allele frequency options (G05, GMF, GOF, GOF*).
• Normalized genomic matrices (GN) to have average diagonal coefficients of 1 and combined with pedigree data.

Main Results:

• Genomic relationship matrices constructed with observed allele frequencies (GOF, GOF*) had smaller average diagonal elements than pedigree-based coefficients.
• Estimates of additive variance were greater in a smaller dataset when genomic matrices deviated from a scale of 1.
• Genomic-assisted EBV accuracies were inflated except when using the normalized genomic matrix (GN).

Conclusions:

• Parameter estimates can be biased if genomic and pedigree relationship matrices are not on a compatible scale.
• Using observed allele frequencies and re-scaling the genomic relationship matrix to an average diagonal of 1 provides reasonable scaling.
• Proper scaling of genomic relationship matrices is vital for unbiased breeding value estimation and accurate accuracy assessment.