Exploratory identification of intestinal health and productive performance patterns in post-weaning piglets using...

Area of Science:

• Animal Science
• Swine Production
• Machine Learning in Agriculture

Background:

• Weaning significantly impacts piglet gut health and performance.
• Intestinal alterations post-weaning are critical challenges in swine production.

Purpose of the Study:

• To apply machine learning for identifying patterns between gut health and productivity in piglets during the early post-weaning phase.
• To explore the potential of explainable AI in animal science for hypothesis generation.

Main Methods:

• Analysis of 103 piglets using 24 histomorphological, biochemical, and productive variables.
• Application of unsupervised (K-means clustering) and supervised (Random Forest, SHAP) machine learning models.
• Identification of key variables differentiating piglet groups based on gut health and growth.

Main Results:

• K-means clustering revealed two distinct groups of piglets with significant differences in intestinal parameters and growth outcomes.
• Villus height-to-crypt depth ratio, absorptive area, maltase activity, and volatile fatty acid concentrations were key differentiating factors.
• Random Forest and SHAP analysis highlighted intestinal morphology, enzymatic activity, and microbial metabolites as strongly associated with cluster classification.

Conclusions:

• Coordinated patterns exist between intestinal function and growth in early post-weaning piglets.
• Identified biomarkers (gut morphology, metabolites) show potential for future nutritional strategies in swine.
• Unsupervised explainable machine learning is a valuable tool for exploratory analysis in animal science research.