Infection profiles in a wild rat–protozoan network are shaped by host traits and environmental factors

Heterogeneity in parasite infection among hosts shapes transmission dynamics and spillover risk to other host species but remains poorly understood in natural systems. We applied network-based stochastic block modeling and machine learning to a uniquely rich dataset to identify and predict protozoan infection profiles in introduced black rats (Rattus rattus) sampled along an environmental gradient in Madagascar. Three host infection profiles emerged, differing in parasite richness and composition, revealing distinct host roles in transmission. Predictive models incorporating host traits (e.g., body mass, microbiome composition) and environmental variables (e.g., population density, habitat structure) accurately classified hosts into profiles, with host traits contributing to predictions 40% more than environmental features. Our findings show how intrinsic and extrinsic factors jointly structure individual-level infection heterogeneity and underscore the value of infection profiles for understanding host–parasite dynamics. Our integrative approach offers a framework for predicting infection risk at human–animal interfaces where zoonotic pathogens circulate.