Search

Arctic charr (Salvelinus alpinus) is a high-value species for the Nordic aquaculture. The highly variable reproductive performance that is commonly observed in commercial farms is hindering the expansion of the Arctic charr industry in Sweden. Traits related to sperm motility (total motility; curvilinear velocity; average path velocity; straight-line velocity) and concentration can play a pivotal role in male fertility. Selective breeding practices could offer solutions and contribute to improving male fertility. The current study aimed to investigate the magnitude of genetic variance for sperm quality traits in a selectively bred population of Arctic charr from Sweden and evaluate the possibility of their improvement through selection. Sperm motility and concentration were recorded using a computer-assisted semen analysis (CASA) system and a NucleoCounter, respectively, in over 400 males from year-class 2017. Double digest restriction-site associated DNA sequencing (ddRAD-seq) was applied in a subset of the recorded animals (n = 329), resulting in the detection of over 5000 single nucleotide polymorphisms (SNPs). Moderate heritability estimates were obtained for the recorded semen traits using both pedigree (0.21–0.32; SE 0.09) and genomic (0.23–0.26; SE 0.09) relationship matrices. A genome-wide association study (GWAS) detected a single SNP significantly associated (P < 1e-05) with total sperm motility on chromosome LG7 in relatively close proximity (500 Kb) to PTPN11a gene previously associated with sperm quality traits in mammals. Moreover, weighted single-step genomic best linear unbiased prediction (WssGBLUP) pinpointed genomic regions explaining more than 3 % of the additive genetic variance for both the motility traits and the sperm concentration. Finally, the efficiency of genomic prediction was tested using a 3-fold cross-validation scheme. Higher prediction accuracy for total motility and velocities (both curvilinear and average path) was obtained using genomic information (0.26–0.29, SE 0.03–0.06) compared to pedigree (0.20–0.28, SE 0.04–0.07), while for sperm concentration a pedigree-based model (0.22 SE 0.03) was more efficient than the genomic model (0.14 SE 0.04). Overall, our results indicate that the recorded sperm quality traits are heritable, and could be improved through selective breeding practices.

•Application of machine learning models for predicting disease resistance in aquaculture.•Extreme gradient boosting was the model that ranked first, demonstrating a slight advantage (1–4%) over GBLUP-MCMC.•All machine learning models had significantly lower computational demands compared to GBLUP-MCMC for threshold traits.