Estimating genomic breeding values from the QTL-MAS Workshop Data using a single SNP and haplotype/IBD approach

Area of Science:

• Quantitative Genetics
• Animal Breeding
• Genomic Selection

Background:

• Genomic selection utilizes high-density markers to predict breeding values, improving accuracy over traditional methods.
• Different statistical models exist for estimating genomic breeding values, each with varying computational demands and biological assumptions.
• The inclusion of polygenic effects alongside genomic information is a key consideration in model development.

Purpose of the Study:

• To compare the accuracy and performance of two genomic selection models: one based on haplotypes/IBD and another on single nucleotide polymorphisms (SNPs).
• To evaluate the impact of including a polygenic effect in genomic models on the accuracy of genomic estimated breeding values (GEBVs) and quantitative trait loci (QTL) detection.
• To assess the computational efficiency and QTL identification capabilities of different genomic prediction approaches.

Main Methods:

• Genomic breeding values were estimated using Gibbs sampling, specifically avoiding the Metropolis-Hastings step.
• Two primary models were applied: a haplotype/IBD approach (20 markers) and a single SNP regression approach.
• Both genomic models were tested with and without the inclusion of a polygenic effect, alongside a control model with only polygenic effects.

Main Results:

• The single SNP approach required significantly fewer effects (11,850) to be estimated compared to the haplotype/IBD approach (366,959).
• Four genomic models identified 11-14 genomic regions with a posterior quantitative trait loci (QTL) probability greater than 0.1.
• Accuracies for GEBVs in later generations ranged from 0.84 to 0.87, with the SNP model showing slightly higher accuracies for total GEBVs.

Conclusions:

• Including a polygenic effect in genomic models did not influence the accuracy of total GEBVs or the prediction of QTL positions.
• The single SNP regression model demonstrated comparable or slightly superior accuracy to the haplotype/IBD model for total GEBVs.
• Both genomic models effectively detected most QTL explaining at least 0.5% of the phenotypic variance, highlighting their utility in genetic improvement.