Introduction
Genomic selection in aquaculture species has been regarded especially useful for hard-to-record-traits , such as disease resistance, product quality and traits that need to be improved in multiple production environments. In the Finnish national breeding programme for rainbow trout, family tanks are used at the initial phase of growth which allows to maintain a pedigree for a large number of fish, and breeding value evaluation is based on this pedigree (PBLUP) (Kause et al. 2022). Genotyping of a portion of the fish accompanied with a single-step genomic evaluation (ssGBLUP) would maintain high selection intensity and simultaneously make use of possibilities of genomic selection. ssGBLUP has been shown as one of the optimal tools, especially when genotyped population is relatively small (Christensen et al., 2012). The combination of pedigree (A) and genomic (G) relationship matrix improves pedigree recording errors and helps utilize sib information more efficiently. The aim of th e current study was 1) to implement ssGBLUP approach for routine evaluation , and 2) to use a validation method to demonstrate the added value of this approach, when breeding for growth and maturity age in two production environments.
Materials and methods
Data from Finnish national breeding programme was obtained from individually tagged fish reared in freshwater (nucleus) and sea stations. Pedigree included 600,409 individuals and 6,234 families made from 3,418 sires and 3,446 dams. Phenotyped fish were born between 1992-2019. Three body weight traits recorded at ages of 2 and 3 years at the nucleus (Weight2, Weight3,) or at age 2 at a sea farm (Sea Weight2) and three binary maturation traits (1=early maturity, 0 = late maturity age) were recorded for males and females at the nucleus (MaturityMale , MaturityFemale ) or at sea (Sea MaturityMale) (Table 1) . Genomic data was available on 4,573 fish born in 2014, 2018, and 2019. Generations 2018 and 2019 were both established from generation 2014, and they had 8,525 and 9,241 tagged fish of which 16% and 28 % were genotyped. Sam ples were genotyped using 57K SNP AxiomTM Trout Genotyping Array. After quality control and imputation 40,374 markers remain for further analysis
Genetic and genomic prediction was performed using mixed-model equations as shown in Kause et al. (2022 ). P edigree BLUP was upgraded to ssGBLUP by replacing A-1 matrix with H-1 computed as:
where G05 is a G matrix constructed with assumption that allele frequency of all markers was equal to 0.5, A22 is a part of A matrix for genotyped fish, w is residual polygenic effect equal to 5% , and st. is a scaling factor equals to .
To test for the predictive ability of PBLUP and ssGBLUP, partial data set was created by removing phenotypes collected in generation 2019 (Table 1) . V alidation of the model fit was done by linear regression of breeding values computed from full ([G]EBV) and partial ([G]EBVpar ) data using formulae (Legarra & Reverter , 2018). Model R2 was assumed as validation reliability.
Results and Discussion
Average v alidation reliability (R2, Table 2) i n ssGBLUP was on 47% (0.18 reliability units) higher than PBLUP implying more accurate prediction. This resulted even thought our validation design was strict - validation group did not share any full- sibs. Sea weight and sea male maturity trait gave lower R2 than freshwater traits due to the lower number of records.
[G]EBVs of v alidation fish were overestimated (b0 >1) in PBLUP and underestimated in ssGBLUP (b0 <1). Dispersion (b1 ) results in PBLUP (<1 ) and ssGBLUP (>1) showed respective under- and overprediction of the [G]EBVs. Low b0 and high b1 in ssGBLUP may be explained by a suboptimal validation design. Removing full year 2019 makes training population small. In such a case n-fold cross-validation might be a better option to use.
Conclusions
A single-step genomic evaluation was successfully implemented in the Finnish national rainbow trout breeding programme . Genotyping of 16-28% tagged fish in the pedigree increased selection accuracy of body weight and maturity age traits.
References
Christensen, O. F., Madsen., P., Nielsen, B., Ostersen, T., Su, G. 2012. Single-step methods for genomic evaluation in pigs. Animal, 6(10), 1565-1571.
Kause A, Nousiainen A, Koskinen H. Improvement in feed efficiency and reduction in nutrient loading from rainbow trout farms: The role of selective breeding. 2022. Journal of Animal Science, 100
Legarra , A., Reverter , A. 2018. Semi-parametric estimates of population accuracy and bias of predictions of breeding values and future phenotypes using the LR method. Genetics Selection Evolution 50:53.