Casein genotypes associate with baseline and dynamic regression components of seminal quality in Murciano-Granadina bucks

Abstract

Abstract This study examined associations between αS1- and κ-casein genotypes and baseline and dynamic regression components of semen quality in Murciano-Granadina bucks using cubic regression models applied to 6,868 ejaculates collected over 10 years, with age at sperm collection as the time axis. Regression decomposition into baseline (b 0 ), linear (b 1 ), curvature (b 2 ), and cubic tren (b 3 ) coefficients enabled quantification of static trait levels and age-dependent dynamics, revealing substantial coefficient variability. For αS1-casein, genotype was associated with baseline progressive motility and total added semen volume ( p  < 0.05), with moderate to large discriminant effects for progressive motility (ηp 2 = 0.277, 95% CI 0.03–0.48) and curvature-related motility and Acrosome Integrity/Intact acrosomes components (ηp 2 = 0.212–0.245), indicating genotype-related differences in modeled semen quality rather than direct fertility effects. In contrast, κ-casein genotypes showed predominantly small to moderate effects, mainly related to ejaculate volume dynamics and baseline sperm concentration (ηp 2 = 0.131–0.138), while most other components were negligible. Cubic terms captured subtle non-linear trajectories, particularly for sperm concentration and endosmosis, highlighting their relevance for long-term trait stability. Discriminant analyses indicated model-based separation of αS1–κ genotype combinations, with EE–AA ( n  = 1) and BE–AA ( n  = 3) aligning with higher sperm motility, concentration, and favorable membrane-integrity responses, whereas BB–BB ( n  = 3) aligned with higher semen volume and lower quality traits. Validation procedures were method-specific. CDA using leave-one-out cross-validation showed stronger discrimination for αS1-casein (74.03%) than κ-casein (64.94%), driven by perfect classification of the dominant BE genotype and substantial misclassification of rarer genotypes, whereas CHAID using 10-fold cross-validation showed moderate predictive performance with lower risk for αS1-casein (0.325, SE = 0.074; accuracy = 67.53%) than κ-casein (0.429, SE = 0.079; accuracy = 57.14%). Press’ Q statistics confirmed that classification accuracy for both genotypes exceeded chance expectation across methods (CDA: αS1-casein Q = 182.40, κ-casein Q = 34.60; CHAID: αS1-casein Q = 143.44, κ-casein Q = 19.64; p  < 0.05), with consistently stronger discrimination for αS1-casein genotypes. Overall, baseline and dynamic regression components capture biologically relevant, exploratory associations between casein polymorphisms and semen quality dynamics, supporting their integration into genomics-informed reproductive management strategies.