Search for a command to run...
Efficient broodstock selection is critical for optimizing hatchery performance in rainbow trout ( Oncorhynchus mykiss) aquaculture, yet reproductive success remains difficult to predict because it emerges from complex interactions among parental physiology, gamete quality, and early offspring performance. This study aimed to develop an integrative, exploratory framework for broodstock evaluation by constructing a Composite Reproductive Index (CRI) that aggregates eyeing, hatching, larval survival, and growth metrics, and by assessing the extent to which pre-spawning parental traits can explain variation in this index. A total of 150 broodstock were allocated to 15 age- and sex-defined sperm–egg treatment groups, with detailed hematological, biochemical, immunological, reproductive, and biometric parameters measured prior to fertilization, followed by standardized spawning and larval rearing. Principal component analysis of ten standardized offspring traits was used to derive the CRI from the first principal component (PC1, 79.17% variance explained), which was rescaled to a 0–100 range. A Random Forest regression model, trained with grouped repeated cross-validation on 25 parental traits, was then used to predict CRI values. The model explained 27.9% of CRI variance (out-of-bag RR), indicating moderate but biologically meaningful explanatory power in a multifactorial reproductive system. Parental age, cholesterol, albumin, AG ratio, leukocyte count, globulin, MCHC, spermatocrit, body length, and hemoglobin emerged as key predictors, highlighting the importance of integrated systemic physiology rather than isolated gamete traits. The highest CRI (100.00) occurred in the Sperm3Egg5 combination, underscoring strong age-related broodstock synergy and substantial heterogeneity among pairings.These findings position the CRI and associated Random Forest model as context-specific, supportive decision tools that can reduce uncertainty in broodstock selection under defined conditions, while emphasizing the need for external validation across independent populations, environments, and species to advance predictive, precision aquaculture. • Novel machine learning model predicts trout reproductive success pre-spawning. • Composite Reproductive Index integrates multi-stage larval performance metrics. • Optimal broodstock age pairing boosts reproductive index by over 70%. • Physiological and immune traits rival gamete metrics in predicting success. • Random Forest modeling enables precision broodstock management in aquaculture.