Reconstructing heterogeneous pathogen interactions from co-occurrence data via statistical network inference
Infectious diseases often involve multiple pathogen species or multiple strains of the same pathogen. As such, knowledge of how different pathogen species or pathogen strains interact is key to understand and predict the outcome of interventions that target only a single pathogen or subset of strains involved in disease. While population-level data have been used to infer pathogen strain interactions, most previously used inference methods only consider uniform interactions between all strains, or focus on marginal interactions between pairs of strains (without correction for indirect interactions through other strains). Here, we evaluate whether statistical network inference could be useful for reconstructing heterogeneous interaction networks from cross-sectional surveys tracking co-occurrence of multi-strain pathogens. To this end, we applied a suite of network models to data simulating endemic infection states of pathogen strains. Satisfactory performance was demonstrated by unbiased estimation of interaction parameters for large sample size. Accurate reconstruction of networks may require regularization or penalizing for sample size. Of note, performance deteriorated in the presence of host heterogeneity, but this could be overcome by correcting for individual-level risk factors. Our work demonstrates how statistical network inference could prove useful for detecting pathogen interactions and may have implications beyond epidemiology.