Bayesian approach integrating correlated foreign information into a multivariate genetic evaluation

Using correlations among traits, two advantages of multivariate genetic evaluations are prediction and improvement of accuracy of estimated breeding values (EBV) for a trait of interest for which phenotypes for some traits of interest, while accurate foreign evaluations for involved correlated traits are routinely performed. Therefore, integration of correlated information provided by foreign evaluations into a local multivariate evaluation could improve its accuracy. The aim of this study was to test a Bayesian approach to combine, through a multivariate evaluation, local pedigree and phenotypes for a trait of interest and genetically correlated foreign information, i.e. EBV and reliabilities provided by a foreign evaluation. One population was simulated across 10 generations. Phenotypes for two traits were simulated for each female. Heritabilities of 0.10 and 0.35 were considered for the trait of interest and the correlated trait, foreign information for sires were locally integrated. No phenotype was locally considered for the correlated trait. For the trait of interest and for correlations from 0.10 to 0.90, results for 100 replicates showed that average rank correlations among Bayesian EBV and EBV from a bivariate evaluation using phenotypes for both traits were close to 1 for sires and ranging from 0.996 to 0.820 for their progeny. The respective correlations for the local evaluation ranged from 0.995 to 0.696 and from 0.993 to 0.651. Mean squared error, expressed as a percentage of the local mean squared error, was lower than 2.76% for sires and 57.78% for progeny. Thereby, the Bayesian approach has the potential to integrate correlated foreign information into a multivariate genetic evaluation.