067
Genomic prediction using a weighted relationship matrix to account for trait architecture in US Holstein cattle
Genomic prediction can be performed through genomic relationship matrix implemented into BLUP (G-BLUP). Assumption of constructing genomic matrix is that contribution is equal for all markers, although weighted trait-specific matrices can be constructed using marker information (effects). Different genomic matrices were constructed using marker information from 3 sources for 9 traits under selection in US Holstein dairy cattle. Results showed that increase in overall predictive ability (higher accuracy and lower bias) can be gained, proportionally to the heritability of the trait. The assumption of complete polygenic architecture is violated in traits with high heritability and presence of quantitative trait loci, and a matrix that takes this into account has been shown to perform better. This method could suit the purpose of increasing accuracy of single-step genomic evaluations.
Keywords:
Genomic relationship matrix
Dairy cattle
Trait genetic architecture