Early-life growth performance and litter characteristics predict gilt selection and first mating success under commercial conditions

Background: Efficient gilt replacement is critical for productivity and sustainability in commercial swine systems. Early-life growth and litter characteristics may influence the likelihood of females progressing through key developmental stages; however, their combined predictive value under commercial conditions remains insufficiently characterised. This study evaluated whether early growth performance and litter-level traits predict progression to the transition phase and attainment of first mating using conventional and machine learning approaches. Methods: A total of 510 female piglets from 77 litters sired by eight boars of the same genetic line were monitored from birth to approximately 220 days of age under commercial farm conditions. Recorded variables included birth weight, transition weight (40–46 days), body weight at ~220 days, and average daily gain (ADG) across growth intervals. Litter characteristics comprised total born, number born alive, number born dead, female proportion, and selection ratio. Variance components attributable to sire and dam were estimated using linear mixed-effects models. Progression to transition (n = 346/520) and attainment of first mating (n = 346 monitored) were analysed using logistic regression and Random Forest models with stratified 10-fold cross-validation. Results: Birth weight was positively correlated with transition weight (r = 0.51, p < 0.01) and early ADG (r = 0.34, p < 0.01), but showed weaker associations with body weight at 220 days (r = 0.17, p < 0.01). Litter size negatively affected birth weight (r = –0.29, p < 0.01) but was not associated with weight at 220 days. Maternal effects accounted for 39.9% of birth weight variance, declining to 10.5% at 220 days, whereas paternal variance remained below 3% across traits (h² = 0.006–0.13). Logistic regression identified birth weight as the strongest predictor of transition (OR = 1.66 per additional kg). Random Forest models demonstrated moderate discriminative ability for transition (AUC = 0.77) and first mating (AUC = 0.74), with early ADG emerging as the most influential predictor of first mating attainment. Conclusions: Early-life growth performance, particularly birth weight and pre-transition ADG, significantly influences gilt developmental progression under commercial conditions. Maternal and litter-level effects explain a substantial proportion of early growth variability. The integration of machine learning approaches improves individual-level prediction beyond traditional regression models and supports data-driven gilt selection strategies aimed at enhancing reproductive efficiency and herd sustainability.