Screening and dynamic change study of microbial and metabolite markers for calf diarrhea based on multi-omics and machine learning

Neonatal calf diarrhea causes substantial early-life mortality and economic losses, yet the dynamic microbiota-metabolite alterations and early-warning biomarkers during disease onset remain poorly defined. Here, we longitudinally profiled fecal microbiota and metabolites in calves from birth to day 20 and integrated machine learning approaches to systematically characterize diarrhea-associated signatures. Diarrheic calves showed reduced α-diversity, and Tyzzerella and Fusobacterium emerged as core differential genera with predictive value validated using an XGBoost model. Differential metabolites were mainly enriched in pathways such as the phosphotransferase system (PTS), and dulcitol and N-acetylmuramate may contribute to diarrhea by modulating intestinal osmolality or inflammatory responses. Notably, a higher abundance of Escherichia-Shigella at birth was potentially associated with subsequent diarrhea risk, while L-glutamic acid, choline, and LysoPC exhibited distinct temporal trajectories. Collectively, these findings provide translational candidate biomarkers to support early warning and microbiota-targeted precision interventions for neonatal calf diarrhea.