Adequacy of Different Pseudo-Phenotypes for Model Training and Validation of Genomic Predictions in Beef Cattle

Abstract Text:

Estimation of marker effects requires an adequate response variable (pseudo-phenotype) to summarize genetic information as well as efficient statistical methods to track the association between pseudo-phenotypes and genotypes of training animals. This study was carried out to investigate the consequences of using three different pseudo-phenotypes obtained after applying different statistical methods for estimation of marker effects in a simulated beef cattle population. Results suggest that deregressed proofs (dEBV) comprise a more suitable response variable to estimate marker effects than estimated breeding values (EBV) or progeny-yield deviations (PYD). When validating genomic predictions, proper inferences about the accuracy and scale of genomic predictions (GEBV) can be obtained using dEBV as target predictand. The accuracy of GEBV should be assessed by the Pearson's correlation between GEBV and dEBV divided by the average accuracy of dEBV in the validation set.

Keywords:

BayesCpi

GBLUP

genomic selection