Leveraging blood biomarkers and machine learning to determine smoltification status of Atlantic salmon (Salmo salar)

Abstract Atlantic salmon ( Salmo salar ) undergoes smoltification, a complex physiological transformation to adapt osmoregulation for life in seawater. Current methods to assess smolt status include visual indices, seawater challenge tests, and gene expression analyses, which are limited in terms of subjectivity, invasiveness, and cost-efficiency. In this study, we developed and validated a machine learning model to evaluate smoltification status using blood biochemical profiles. Atlantic salmon juveniles were reared under controlled conditions and sampled at 11 time points. Blood electrolytes and gene expression of Na⁺/K⁺-ATPase isoforms NKAα1a and NKAα1b were analyzed, alongside seawater tolerance through plasma chloride levels. Using XGBoost, a state-of-the-art gradient boosting algorithm, we achieved high classification accuracy (ROC AUC = 0.99) in distinguishing smolt from non-smolt individuals. Plasma chloride, calcium, and sodium were the most predictive features, while glucose, potassium, and creatinine also provided predictive value. Regression models predicting DDCt qPCR indices (Spearman r = 0.89) and seawater plasma chloride (Spearman r = 0.78) both demonstrated good performance. These findings suggest that blood-based point-of-care testing (POCT) combined with machine learning offers a rapid, reliable, and less invasive method to assess smoltification, with potential to replace traditional seawater challenges and molecular assays in hatchery settings.