BackgroundSalmon Rickettsial Syndrome (SRS) caused by Piscirickettsia salmonis is a major disease affecting the Chilean salmon industry. Genomic selection (GS) is a method wherein genome-wide markers and phenotype information of full-sibs are used to predict genomic EBV (GEBV) of selection candidates and is expected to have increased accuracy and response to selection over traditional pedigree based Best Linear Unbiased Prediction (PBLUP). Widely used GS methods such as genomic BLUP (GBLUP), SNPBLUP, Bayes C and Bayesian Lasso may perform differently with respect to accuracy of GEBV prediction. Our aim was to compare the accuracy, in terms of reliability of genome-enabled prediction, from different GS methods with PBLUP for resistance to SRS in an Atlantic salmon breeding program. Number of days to death (DAYS), binary survival status (STATUS) phenotypes, and 50 K SNP array genotypes were obtained from 2601 smolts challenged with P. salmonis. The reliability of different GS methods at different SNP densities with and without pedigree were compared to PBLUP using a five-fold cross validation scheme.ResultsHeritability estimated from GS methods was significantly higher than PBLUP. Pearson’s correlation between predicted GEBV from PBLUP and GS models ranged from 0.79 to 0.91 and 0.79–0.95 for DAYS and STATUS, respectively. The relative increase in reliability from different GS methods for DAYS and STATUS with 50 K SNP ranged from 8 to 25% and 27–30%, respectively. All GS methods outperformed PBLUP at all marker densities. DAYS and STATUS showed superior reliability over PBLUP even at the lowest marker density of 3 K and 500 SNP, respectively. 20 K SNP showed close to maximal reliability for both traits with little improvement using higher densities.ConclusionsThese results indicate that genomic predictions can accelerate genetic progress for SRS resistance in Atlantic salmon and implementation of this approach will contribute to the control of SRS in Chile. We recommend GBLUP for routine GS evaluation because this method is computationally faster and the results are very similar with other GS methods. The use of lower density SNP or the combination of low density SNP and an imputation strategy may help to reduce genotyping costs without compromising gain in reliability.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-017-3487-y) contains supplementary material, which is available to authorized users.

Salmonid rickettsial syndrome (SRS), caused by the intracellular bacterium Piscirickettsia salmonis, is one of the main diseases affecting rainbow trout (Oncorhynchus mykiss) farming. To accelerate genetic progress, genomic selection methods can be used as an effective approach to control the disease. The aims of this study were: (i) to compare the accuracy of estimated breeding values using pedigree-based best linear unbiased prediction (PBLUP) with genomic BLUP (GBLUP), single-step GBLUP (ssGBLUP), Bayes C, and Bayesian Lasso (LASSO); and (ii) to test the accuracy of genomic prediction and PBLUP using different marker densities (0.5, 3, 10, 20, and 27 K) for resistance against P. salmonis in rainbow trout. Phenotypes were recorded as number of days to death (DD) and binary survival (BS) from 2416 fish challenged with P. salmonis. A total of 1934 fish were genotyped using a 57 K single-nucleotide polymorphism (SNP) array. All genomic prediction methods achieved higher accuracies than PBLUP. The relative increase in accuracy for different genomic models ranged from 28 to 41% for both DD and BS at 27 K SNP. Between different genomic models, the highest relative increase in accuracy was obtained with Bayes C (∼40%), where 3 K SNP was enough to achieve a similar accuracy to that of the 27 K SNP for both traits. For resistance against P. salmonis in rainbow trout, we showed that genomic predictions using GBLUP, ssGBLUP, Bayes C, and LASSO can increase accuracy compared with PBLUP. Moreover, it is possible to use relatively low-density SNP panels for genomic prediction without compromising accuracy predictions for resistance against P. salmonis in rainbow trout.

Sea lice infestations caused by Caligus rogercresseyi are a main concern to the salmon farming industry due to associated economic losses. Resistance to this parasite was shown to have low to moderate genetic variation and its genetic architecture was suggested to be polygenic. The aim of this study was to compare accuracies of breeding value predictions obtained with pedigree-based best linear unbiased prediction (P-BLUP) methodology against different genomic prediction approaches: genomic BLUP (G-BLUP), Bayesian Lasso, and Bayes C. To achieve this, 2404 individuals from 118 families were measured for C. rogercresseyi count after a challenge and genotyped using 37 K single nucleotide polymorphisms. Accuracies were assessed using fivefold cross-validation and SNP densities of 0.5, 1, 5, 10, 25 and 37 K. Accuracy of genomic predictions increased with increasing SNP density and was higher than pedigree-based BLUP predictions by up to 22%. Both Bayesian and G-BLUP methods can predict breeding values with higher accuracies than pedigree-based BLUP, however, G-BLUP may be the preferred method because of reduced computation time and ease of implementation. A relatively low marker density (i.e. 10 K) is sufficient for maximal increase in accuracy when using G-BLUP or Bayesian methods for genomic prediction of C. rogercresseyi resistance in Atlantic salmon.

The objectives of this study were: (i) to assess genotype imputation accuracy in different scenarios using genomewide single nucleotide polymorphisms (SNP) data from a population comprising two generations of farmed Atlantic salmon and (ii) to assess the accuracy of genomic predictions for a quantitative trait (body weight) using the imputed genotypes. The pedigree consisted of 53 parents and 1069 offspring genotyped using a high-density SNP panel (50 K). Two groups were created: Group A: 90% of the offspring were included into training and 10% into validation sets; Group B: 10% of the offspring were included into training and 90% into validation sets. Different scenarios of available genotypic information from relatives were tested for the two groups previously described. Imputation was performed using three in silico low-density panels (0.5, 3 and 6 K) with all markers except the markers present on the low-density panel masked in the validation sets. The accuracy of genomic selection was tested using the scenarios that resulted in the best and the worst imputation accuracy for the three low density panels and were compared to accuracy obtained from pedigree-based best linear unbiased prediction (PBLUP) and genomic predictions using the 50 K SNP panel. In general, imputation accuracy ranged from 0.74 to 0.98 depending on scenario. For the best scenario with the highest number of animals in reference population (Group A), the accuracy of imputation ranged from 0.95 to 0.98 depending on the low-density panel used. For the best scenario with the lowest number of animals in reference population (Group B), the accuracy of imputation ranged from 0.94 to 0.98 depending on the low-density panel used. In general, the number of SNPs in the low-density panels had a greater influence on the accuracy of imputation than the size of the reference set. The accuracies of genomic predictions using imputed genotypes, ranging from 0.71 to 0.73, outperformed PBLUP (0.66) and were identical or very similar to the use of all true genotype data (0.73). The high imputation and genomic prediction accuracy suggest that the imputation of genotypes from low density (0.5 to 3 K) to high density (50 K) could be a cost-effective strategy for the feasibility of the practical implementation of genomic selection in Atlantic salmon.