Multi-Omics and Machine Learning for Predicting Foodborne Pathogen Carriage in Food-Producing Animals: Current Knowledge, Challenges, and Future Applications

Food-producing animals are important reservoirs of foodborne pathogens, but asymptomatic carriage and intermittent shedding complicate reliable risk stratification. Multi-omics and machine learning provide complementary approaches for investigating carriage as a systems-level phenotype shaped by host susceptibility, resident microbiota, pathogen traits, functional activity, and production conditions. This review evaluates current evidence for omics-based and machine-learning approaches to understand, classify, and predict pathogen carriage in major food-producing animal systems. Evidence is strongest in poultry, particularly in Salmonella-focused research, whereas swine and cattle studies provide important microbiome, genomic, and host-response biomarkers, and evidence in small ruminants and aquaculture remains limited. Across animal–pathogen systems, association, biomarker-discovery, and concurrent-classification studies predominate, while prospective prediction and independent external validation remain uncommon. Direct head-to-head comparisons of multi-omics with matched single-omics or conventional approaches are scarce, leaving their incremental predictive value uncertain. Key barriers include inconsistent outcomes, intermittent shedding, limited biological replication, technical heterogeneity, overfitting, and insufficient external validation. Near-term applications are most plausible for risk-based surveillance, targeted sampling, group-level prioritization, and environmental early warning.