571. Genomic prediction of complex traits in European sea bass: random cross-validation and leave-one-family-out validation

Abstract

For European sea bass (Dicentrarchus labrax L.) hatcheries, viral nervous necrosis mortality, post-stress cortisol concentration, antibody titer against nervous necrosis virus and body weight are interesting traits to be included in the breeding objectives. An experimental population (n=650) generated using a commercial broodstock was phenotyped for those traits and genotyped with a genome-wide SNP panel (16,075 markers). The accuracy of the genomic model in the prediction of estimated breeding values was assessed in a 5-fold random cross-validation and in a leave-one-family-out scenario, where the prediction equation was obtained by training the model on the estimated breeding values of all animals except the offspring of the parent used as test set. The accuracies (Pearson’s r) were different across traits and equal to 0.90 (random cross-validation) and 0.59 (leave-one-family-out) for mortality, 0.86 and 0.60 for cortisol concentration, 0.79 and 0.01 for antibody titre, and 0.71 and 0.18 for body weight.