PS10-5. Predictive Modeling for Reproductive Potential of Gilts Using Vaginal Microbiomes.

Abstract While reproductive potential is essential for animal production and economic success, the low heritability of fertility and litter-related traits can limit selection efficiency. The vaginal microbiome plays an important role in maintaining reproductive health, and recent research has begun to explore the potential relationship between vaginal microorganisms and fertility. Yet it is currently unknown whether the vaginal microbiome can be used to develop predictive models for reproductive outcomes. This study aimed to evaluate whether the vaginal microbiome composition in gilts, combined with machine learning approaches, could be used to predict reproductive outcomes. Vaginal microbiome samples were collected from gilts at approximately 190 days of age, prior to their first breeding. Based on subsequent reproductive performance, animals were categorized as either Fertile n = 30 or being infertile n = 13. Bacterial DNA was sequenced using 16S rRNA gene amplification and long-read Nanopore sequencing to determine Species-level relative abundances. Feature selection was then conducted using Partial Least Squares Discriminant Analysis (PLS-DA) and Recursive Feature Elimination (RFE). The selected genera were used to train and evaluate five machine learning classifiers using repeated stratified cross-validation. All models achieved promising predictive performance during validation and testing, with the Logistic Regression classifier achieving the best performance, correctly predicting Fertile gilts with 92 ± 2% accuracy (SE) and infertile gilts 82 ± 3% accuracy (SE) across 100 unique train–test splits. These findings suggest that the vaginal microbiome represents a promising source of predictors for reproductive performance in gilts and may contribute to the development of new methods for reproductive selection in swine breeding. Future work will focus on expanding the dataset and further validating the prediction models to enhance their robustness and practical applicability.