Optimizing genomic prediction model given causal genes in a dairy cattle population.

Area of Science:

• Quantitative Genetics
• Animal Breeding
• Genomics

Background:

• Genomic prediction (GP) accuracy shows limited gains with whole-genome sequence data compared to SNP chips.
• Genome-wide association studies (GWAS) identify genomic regions and potential causal genes, offering valuable prior knowledge for GP.
• Previous GP model modifications incorporating prior knowledge have had limited validation across diverse genetic architectures.

Purpose of the Study:

• To evaluate GP performance across varied genetic architectures using known causal genes.
• To compare modified GP models that emphasize causal genes and explore different weighting strategies for these genes.

Main Methods:

• Simulated pseudo-phenotypes based on real dairy cattle genotypes and phenotypes.
• Evaluated classical genomic best linear unbiased prediction (GBLUP) and three modified GP models.
• Modified models incorporated causal genes as fixed effects, a separate random component, or within the genomic relationship matrix.

Main Results:

• Highlighting known causal genes, which explained significant genetic variance, increased predictive accuracy.
• The optimal strategy involved combining all known causal genes into the genomic relationship matrix.
• Treating causal genes as a separate random component was effective when they explained >20% of genetic variance.
• Differential weighting of causal genes further enhanced predictive accuracy.

Conclusions:

• Incorporating known causal genes into GP models substantially improves predictive accuracy.
• The genomic relationship matrix approach is superior for integrating causal genes in GP.
• Strategic weighting and modeling of causal genes are crucial for maximizing GP performance in complex traits.