Abstract
Simple SummarySelective breeding programs based on genomic data are still not a common practice in aquaculture, although genomic selection has been widely demonstrated to be advantageous when trait phenotyping is a difficult task. In this study, we investigated the accuracy of predicting the phenotype and the estimated breeding value (EBV) of three Bayesian models and a Random Forest algorithm exploiting the information of a genome-wide SNP panel for European sea bass. The genomic predictions were developed for mortality caused by viral nervous necrosis, post-stress cortisol concentration, antibody titer against nervous necrosis virus and body weight. Selective breeding based on genomic data is a possible option for improving these traits while overcoming difficulties related to individual phenotyping of the investigated traits. Our results evidenced that the EBV used as a pseudo-phenotype enhances the predictive performances of genomic models, and that EBV can be predicted with satisfactory accuracy. The genomic prediction of the EBV for mortality might also be used to classify the phenotype for the same trait.In European sea bass (Dicentrarchus labrax L.), the viral nervous necrosis mortality (MORT), post-stress cortisol concentration (HC), antibody titer (AT) against nervous necrosis virus and body weight (BW) show significant heritability, which makes selective breeding a possible option for their improvement. An experimental population (N = 650) generated by a commercial broodstock was phenotyped for the aforementioned traits and genotyped with a genome-wide SNP panel (16,075 markers). We compared the predictive accuracies of three Bayesian models (Bayes B, Bayes C and Bayesian Ridge Regression) and a machine-learning method (Random Forest). The prediction accuracy of the EBV for MORT was approximately 0.90, whereas the prediction accuracies of the EBV and the phenotype were 0.86 and 0.21 for HC, 0.79 and 0.26 for AT and 0.71 and 0.38 for BW. The genomic prediction of the EBV for MORT used to classify the phenotype for the same trait showed moderate classification performance. Genome-wide association studies confirmed the polygenic nature of MORT and demonstrated a complex genetic structure for HC and AT. Genomic predictions of the EBV for MORT could potentially be used to classify the phenotype of the same trait, though further investigations on a larger experimental population are needed.
Highlights
In fish species, improvements in selection accuracy due to genomic predictions and benefits of genomic selection over pedigree-based methods have been documented both in simulation studies [1] and real data from applied selective programs for growth and disease resistance [2,3,4,5,6,7]
This study aimed to investigate the performance of genomic models by making use of the individual genotypes of 16,075 genome-wide SNPs as predictors of the viral nervous necrosis (VNN) mortality, post-stress cortisol concentration, nervous necrosis virus (NNV) antibody titer and body weight, predicting both trait phenotypes and pseudo-phenotypes (EBV)
Since moderate genetic correlations between MORT and body weight (BW) or antibody titer (AT) were estimated in a previous study [13] on the same data, we investigated the performance of the genomic predictions of the EBVFULL for traits correlated with MORT when used as classifiers of the observed mortality
Summary
Improvements in selection accuracy due to genomic predictions and benefits of genomic selection over pedigree-based methods have been documented both in simulation studies [1] and real data from applied selective programs for growth and disease resistance [2,3,4,5,6,7]. Disease outbreaks may have important impacts on aquaculture, threatening the survival of farmed animals, interfering with the productivity and sustainability of intensive systems and causing significant economic losses. Selective breeding applied to aquaculture for economically important traits is a relatively young practice compared to plant and livestock species [8], and its development has mainly followed the industrialization and growth of the aquaculture sector [9,10]. Beyond biosecurity and management practices, selective breeding for disease resistance is viewed as a feasible and sustainable approach to prevent and control damages inflicted by disease outbreaks in aquaculture farms [8,11]. Low-to-moderate heritability (h2) estimates for VNN resistance [2,13,14,15] indicate the presence of a genetic basis for this trait, and opportunities to develop selective breeding approaches
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
