Defining a Parameter Space for GMACE

Abstract

National genomic evaluations of young bulls (GEBV) are combined by Interbull, using an international genomic MACE model (GMACE), with non-zero residual correlations to account for sharing of genotypes among national genomic evaluation systems.  It was observed recently that GMACE results for mastitis resistance were inconsistent with corresponding results for somatic cell score.  This study examined the current GMACE methods, and new modifications to better account for different heritabilities and for different genomic reliabilities among countries for a given trait.  A parameter space was defined that bounds GMACE results, on the scale of each country, to fall somewhere between the national GEBV, and predictions of international GEBV when sharing of genotypes, common SNP panels, etc, are ignored.  Distances to either boundary were estimated as a function of the degree of data sharing observed among national genomic evaluation systems.  The proposed modifications to GMACE had largest effects on traits with a wide range of heritabilities among countries, such as mastitis resistance, and on the scales of countries that had relatively low national genomic reliabilities.  Results from GMACE were much more consistent between mastitis and somatic cell score after the modifications, and also among all other traits evaluated by Interbull.