An approach for diagnosis of diarrhea in neonatal piglets based on the core gut microbiota and machine learning.

Diarrheal diseases, such as yellow dysentery and white dysentery caused by pathogens or viruses, in newborn piglets lead to substantial economic losses in the swine industry worldwide. Gut microbiota dysbiosis is frequently observed in diarrheic piglets and is thought to play a role in disease pathogenesis, although causal relationships remain to be established. However, developing reliable microbiome-based diagnostic tools still poses a significant challenge. This study aimed to develop a diagnostic model for piglet diarrhea by integrating core microbiota analysis with machine learning. Fecal samples from diarrheic and healthy piglets were subjected to metagenomic sequencing to characterize archaeal, bacterial, and fungal communities. We identified diarrhea-associated bacterial biomarkers via LEfSe, DESeq2, and microbial cooccurrence network analysis. These microbial features were used to construct and compare multiple machine learning classifiers. Our results revealed significant disparities in the structure and diversity of the gut microbiota between diarrheic and healthy piglets, with the bacterial community showing the most notable changes. Among the models developed, the decision tree classifier based on bacterial genus-level features achieved the highest prediction accuracy of 91.18%. Furthermore, a simplified model utilizing a panel of 18 core bacterial genera also demonstrated high efficacy, with a support vector machine model achieving 88.24% accuracy. In independent validation using our internal dataset, the random forest model exhibited the best generalizability and stability. This study establishes a robust, microbiota-based diagnostic model for diarrhea in neonatal piglets, highlighting the potential of machine learning in leveraging microbiome data for disease classification and health management in livestock production.