Effect of different genomic relationship matrices on accuracy and scale

Area of Science:

• Animal Genetics and Breeding
• Quantitative Genetics
• Genomic Selection

Background:

• Accurate genetic evaluation is crucial for broiler breeding programs.
• Single-step genomic best linear unbiased prediction (ssGBLUP) integrates pedigree and genomic data.
• The construction of the genomic relationship matrix (G) impacts ssGBLUP accuracy.

Purpose of the Study:

• To compare different methods of constructing the genomic relationship matrix (G) within ssGBLUP.
• To evaluate the impact of minor allele frequency (MAF) thresholds on G matrix properties and predictive ability.
• To determine optimal strategies for unbiased genomic evaluation in broilers.

Main Methods:

• Phenotypic data (body weight, breast meat area) from 287,614 broilers analyzed.
• Genotyping of 4,113 birds for 57,636 SNPs.
• ssGBLUP implemented using G matrices constructed with equal (GEq) or current (GC) allele frequencies, with varying MAF thresholds, and compared with pedigree matrix A(22).

Main Results:

• GC matrix construction with MAF thresholds resulted in average diagonal and off-diagonal elements compatible with A(22).
• Predictive ability was similar across methods, with minor variations at high MAF thresholds.
• ssGBLUP using GEq resulted in upward bias in estimated breeding values (EBVs), while GC resulted in downward bias; biases were eliminated by scaling GC.

Conclusions:

• Scaling the GC matrix to be compatible with A(22) provides unbiased genomic evaluations in ssGBLUP.
• Reducing SNP with small MAF has minimal impact on real accuracy but can inflate estimated accuracies.
• Optimized G matrix construction is essential for accurate and unbiased genomic selection in broiler breeding.