Genomic selection models double the accuracy of predicted breeding values for bacterial cold water disease resistance compared to a traditional pedigree-based model in rainbow trout aquaculture

Abstract

BackgroundPreviously, we have shown that bacterial cold water disease (BCWD) resistance in rainbow trout can be improved using traditional family-based selection, but progress has been limited to exploiting only between-family genetic variation. Genomic selection (GS) is a new alternative that enables exploitation of within-family genetic variation.MethodsWe compared three GS models [single-step genomic best linear unbiased prediction (ssGBLUP), weighted ssGBLUP (wssGBLUP), and BayesB] to predict genomic-enabled breeding values (GEBV) for BCWD resistance in a commercial rainbow trout population, and compared the accuracy of GEBV to traditional estimates of breeding values (EBV) from a pedigree-based BLUP (P-BLUP) model. We also assessed the impact of sampling design on the accuracy of GEBV predictions. For these comparisons, we used BCWD survival phenotypes recorded on 7893 fish from 102 families, of which 1473 fish from 50 families had genotypes [57 K single nucleotide polymorphism (SNP) array]. Naïve siblings of the training fish (n = 930 testing fish) were genotyped to predict their GEBV and mated to produce 138 progeny testing families. In the following generation, 9968 progeny were phenotyped to empirically assess the accuracy of GEBV predictions made on their non-phenotyped parents.ResultsThe accuracy of GEBV from all tested GS models were substantially higher than the P-BLUP model EBV. The highest increase in accuracy relative to the P-BLUP model was achieved with BayesB (97.2 to 108.8%), followed by wssGBLUP at iteration 2 (94.4 to 97.1%) and 3 (88.9 to 91.2%) and ssGBLUP (83.3 to 85.3%). Reducing the training sample size to n = ~1000 had no negative impact on the accuracy (0.67 to 0.72), but with n = ~500 the accuracy dropped to 0.53 to 0.61 if the training and testing fish were full-sibs, and even substantially lower, to 0.22 to 0.25, when they were not full-sibs.ConclusionsUsing progeny performance data, we showed that the accuracy of genomic predictions is substantially higher than estimates obtained from the traditional pedigree-based BLUP model for BCWD resistance. Overall, we found that using a much smaller training sample size compared to similar studies in livestock, GS can substantially improve the selection accuracy and genetic gains for this trait in a commercial rainbow trout breeding population.Electronic supplementary materialThe online version of this article (doi:10.1186/s12711-017-0293-6) contains supplementary material, which is available to authorized users.

Potential of genomic selection for improvement of resistance to Ostreid Herpes virus in Pacific oyster (Crassostrea gigas)

“…The prediction accuracy values ranged from 0.678 to 0.758 for GBLUP (with SNP densities ranging from 500 to 18K), while PBLUP only reached an accuracy of 0.637. This result has been mirrored in other studies of genomic versus pedigree-based prediction of disease resistance breeding values for other important farmed fish species, e.g Atlantic salmon (Tsai et al 2015; Yoshida et al 2017; Ødegård et al 2014; Robledo et al 2018), rainbow trout (Vallejo et al 2017; Yoshida et al 2018), sea bream (Palaiokostas et al 2016) and sea bass (Palaiokostas et al 2018a). Further, in shellfish similar findings have been observed for prediction of breeding values for growth traits in scallop (Dou et al 2016) and Pacific oyster (Gutierrez et al 2018b).…”

Potential of genomic selection for improvement of resistance to ostreid herpesvirus in Pacific oyster (Crassostrea gigas)

“…This result has been mirrored in other studies of genomic vs. pedigree‐based prediction of disease resistance breeding values for other important farmed fish species, e.g. Atlantic salmon (Ødegård et al ; Tsai et al ; Yoshida et al ; Robledo et al ), rainbow trout (Vallejo et al ; Yoshida et al ), sea bream (Palaiokostas et al ) and sea bass (Palaiokostas et al ). Further, in shellfish similar findings have been observed for prediction of breeding values for growth traits in scallop (Dou et al ) and Pacific oyster (Gutierrez et al ).…”

Population genomic structure and genome-wide linkage disequilibrium in farmed Atlantic salmon (Salmo salarL.) using dense SNP genotypes

“…A r 2 threshold of 0.2 was estimated for marker pairs separated by 8,000 Kb,42 45 and 64 Kb in the NAM, NOR and SCO populations, respectively. In this study we show 46 that this SNP panel can be used to detect association between markers and traits of 47 interests and also to capture high-resolution information for genome-enabled 48 estimated breeding values (GEBVs) in farmed salmon species ; 82 Barría et al, 2018;Ødegård et al, 2014;Tsai et al, 2016;Vallejo et al, 2017;Yoshida 83 et al, 2018). Furthermore, association mapping through genome wide association 84 studies (GWAs) is a useful approach to detect genomic regions and genes involved in 85 economically important traits for salmon aquaculture and they also rely on LD between 86 the QTL and SNP markers.…”