Genomic prediction has been increasingly applied in aquaculture selective breeding; however, systematic evaluations of prediction accuracy across multiple aquaculture species and analytical methods under a unified and comparable framework remain limited. In this study, we conducted a comprehensive comparative assessment of genomic prediction performance across four representative aquaculture species, including Atlantic salmon (Salmo salar), gilthead sea bream (Sparus aurata), common carp (Cyprinus carpio), and rainbow trout (Oncorhynchus mykiss), using ten genomic prediction models including GBLUP, Bayesian and machine learning methods. Prediction accuracy varied widely among species and models, ranging from 0.49 to 0.85, and was strongly associated with trait heritability. High-heritability traits consistently achieved higher prediction accuracies, with rainbow trout and common carp exhibiting the best overall performance (0.75–0.83 and 0.73–0.85, respectively), whereas Atlantic salmon and gilthead sea bream showed lower and more variable accuracies (0.49–0.61 and 0.49–0.66). No single model performed optimally across all species. Machine learning-based approaches achieved the highest prediction accuracy in specific cases but exhibited pronounced species-dependent variability, while GBLUP provided stable and well-calibrated predictions with consistently low bias. Incremental SNP feature selection further improved prediction accuracy by 2.8–4.2% in three species using only 0.54–9.64% of the available markers, whereas no improvement was observed for a low-heritability trait. These results show that genomic prediction performance is highly context-dependent and underscores the importance of jointly considering trait genetic architecture, population characteristics, model choice, and marker selection when optimizing genomic selection strategies in aquaculture breeding programs.

Introduction: Recirculating aquaculture systems (RAS) are increasingly used in intensive aquaculture because of their capacity to reduce nitrogen, phosphorus, and organic matter discharges. However, whether improved conventional effluent quality is accompanied by a reduced antibiotic resistance gene (ARG) burden remains unclear. Methods: During the mid-cultivation stage, effluents were collected from four RAS farms and four paved flow-through raceway systems (PFRS). Water quality parameters, including total nitrogen (TN), total phosphorus (TP), chemical oxygen demand (COD), chlorophyll-a (Chl-a), and dissolved oxygen (DO), were measured. Microbial community composition was characterized using 16S rRNA gene sequencing, while ARGs and mobile genetic elements (MGEs) were quantified using GeoChip 5.0. Multivariate statistics and co-occurrence network analyses were used to evaluate associations among environmental conditions, microbial communities, ARG profiles, and MGE profiles. Results: RAS effluents showed substantially lower nutrient and organic loads than PFRS, confirming improved conventional water quality. However, total ARG abundance was significantly higher in RAS than in PFRS, increasing from 35.39 ± 11.43 × 10 ⁶ to 63.99 ± 15.33 × 10 ⁶ , corresponding to a 1.81-fold increase. In contrast, total MGE abundance was numerically higher in RAS but did not differ significantly between systems. Microbial alpha diversity was lower in RAS, and beta diversity analysis showed distinct community structuring between cultivation modes, with both compositional differences and dispersion effects contributing to group separation. ARG class- and subtype-level profiles showed heterogeneous abundance patterns, with several major ARG categories remaining highly abundant in RAS, although most feature-level differences were not significant after FDR correction. Co-occurrence network analysis further revealed structured associations among microbial species, ARGs, and MGEs, with modular organization and highly connected resistance-associated nodes. Conclusion: This study demonstrates a “low-pollution but non-reduced ARG burden” pattern in RAS effluents. Improved removal of conventional pollutants did not correspond to a proportional reduction in ARG or MGE profiles. These findings indicate that conventional water-quality indicators alone are insufficient to evaluate resistance-associated genetic pollution in engineered aquaculture systems, and that ARGs, MGEs, microbial community structure, and species–ARG/MGE association patterns should be integrated into future RAS monitoring and management.

Genotype imputation is a cost-effective strategy for achieving dense genome-wide genotyping, but its accuracy in rainbow trout (Oncorhynchus mykiss) has been hampered by the lack of a comprehensive, multi-population genome-wide variant and imputation resource. Here, we present TroutHap, a multi-population genome-wide variant and imputation resource constructed from 540 high-coverage whole-genome sequences (~12.58×) representing multiple geographic sources and breeding backgrounds, mainly from North America, Europe, and China. We identified 16,408,551 high-confidence SNPs and integrated them into a unified genome-wide variant map across all 30 chromosome pairs. Using this resource, we benchmarked genotype imputation across simulated low-coverage sequencing depths. At 1× sequencing depth, genotype accuracy achieves >90%, while additional sequencing beyond 2× yielded smaller improvements. Supplementary cross-platform validation using a public 57 K SNP array dataset further evaluated compatibility with array-based genotyping data, with genotype concordance and genotype accuracy reaching 80.5% and 71.9%, respectively, after site-level filtering. TroutHap is released as an open community resource, including high-confidence SNP site lists, representative chromosome-level variant files, representative imputed genotype files, metadata tables, validation summaries, and reproducible workflow outputs. This dataset provides a reusable resource for low-coverage sequencing imputation, SNP-array-based genotype harmonization, and downstream genomic analyses in rainbow trout.