Aquaculture has great potential to enhance food security and meet the increasing consumer demand for seafood [1, 2]. However, one of the challenges is the lack of genetically improved strains of fish for aquaculture [3, 4]. Selective breeding programs can produce animals of improved genetics for heritable traits that positively impact aquaculture production [5, 6]. Breeding programs in rainbow trout have focused on the growth rate, disease resistance, and fat content [7–9].
Muscle yield has been ranked among the top traits impacting the industry returns . Also, quality attributes can affect industry profitability and determine consumer’s attitudes towards the product. For instance, loss of fillet firmness contributes to fillet downgrading and economic losses to the industry . Even though important for the industry, muscle yield and firmness have not received much attention because they cannot be measured directly on breeding candidates, which makes genetic selection for these traits hard to implement [12, 13]. Moderate levels of heritability estimates for muscle yield  and fillet firmness  have been reported in rainbow trout, allowing potential genetic improvement through selective breeding programs .
Statistical models based on phenotypes and pedigree information have been widely used in traditional genetic improvement programs to estimate EBV and identify the best selection candidates in animal populations . However, applying genomic approaches has the potential to enhance and expedite genetic gains in breeding programs . Although the implementation of genomic selection (GS) in livestock animals started in 2008 , it took more time to have this technology adopted in aquaculture species. The delayed incorporation of genomic information in rainbow trout breeding programs was mainly due to the lack of dense SNP arrays . A recently developed 50K SNP chip revealed the complex polygenic nature of muscle yield  and fillet firmness , suggesting GS as a practical strategy for rainbow trout breeding.
Selection based on GEBV can be performed at an early age with high accuracy, once a DNA sample for selection candidates is obtained . In agricultural livestock species, the use of genomic information has been shown to be effective in increasing accuracy of GEBV. For instance, Garcia-Ruiz et al.  reported an increase of at least 50% in the rate of genetic gain for several traits in US Holstein cattle. Besides, the gains in accuracy of GEBV with the implementation of GS were assessed for harvest and carcass weight, and growth and fillet yield in catfish , and tilapia , respectively. For rainbow trout, the accuracy of GEBV was assessed for body weight, carcass weight, fillet weight, fillet yield , and resistance to diseases such as bacterial cold water disease (BCWD) [24, 25], columnaris disease , and infectious pancreatic necrosis virus . GS will allow within-family selection and hence increase the accuracy of genomic predictions and selection response, especially for lethally measured traits . Additionally, GS has the potential to decrease the rates of inbreeding by selecting non-sib candidates from more families . Genomic prediction can lead to higher genetic gains, relative to pedigree-based selective breeding methods, which may, partially, cover the extra cost of genotyping . Furthermore, GS is particularly advantageous in aquaculture species because the high fecundities of these species allow for the rapid amplification of elite genetics.
Although the use of genomic information generated from high-density SNP arrays has been demonstrated to expedite the rate of genetic gain in breeding programs, the genotyping cost is still high and alternative strategies are needed to reduce the cost of identifying elite breeding candidates . Cost-effective strategies were previously assessed, yielding higher genomic prediction accuracies than those estimated using the pedigree-based PBLUP model [23, 24, 29]. The cost-reducing methods include using reduced-density SNP panels [24, 30, 31] and genotype imputation [20, 23, 32, 33]. However, imputations are prone to errors, which leads to less reliable genomic predictions . Recently, the impact of low-density SNP panels on the accuracy of genomic predictions has been increasingly studied [24, 30, 31, 34, 35]. In rainbow trout, a 500 reduced-density SNP panel showed higher genomic prediction accuracies for BCWD resistance (0.50–0.56) than traditional EBV (0.36) . The study showed that the ssGBLUP model outperformed other genomic models, yielding high accuracy of genomic predictions with the least bias when reducing SNP panel density .
Recently, the accuracy of genomic prediction was assessed for muscle yield in rainbow trout using a 57K genomic SNP panel ; however, the impact of reducing the SNP panel density on the gains in accuracy for muscle yield relative to the traditional PBLUP has not been investigated. Additionally, there are no reports on the benefits of using GS for quality traits such as fillet firmness in rainbow trout. Therefore, the objectives of this study were to utilize a recently developed 50K transcribed SNP chip  to 1) evaluate the predictive ability of GS for muscle yield and fillet firmness in rainbow trout; and 2) investigate the impact of reducing the SNP panel density on the ability to predict muscle yield and fillet firmness.