Reconstructing multi-strain pathogen interactions from cross-sectional survey data via statistical network inference

Area of Science:

• Epidemiology
• Computational Biology
• Infectious Disease Dynamics

Background:

• Infectious diseases frequently involve multiple pathogen species or strains.
• Existing methods for inferring pathogen interactions are limited, often overlooking indirect effects and leading to biased results.
• Accurate understanding of pathogen interactions is vital for effective disease intervention strategies.

Purpose of the Study:

• To evaluate statistical network inference for reconstructing heterogeneous interactions among multiple pathogen strains.
• To assess the ability of these methods to detect joint presence/absence patterns of pathogen strains within hosts using cross-sectional survey data.

Main Methods:

• Applied various network models to simulated survey data representing endemic infection states with potential interactions.
• Investigated the impact of regularization and penalization techniques for sample size on interaction network reconstruction.
• Assessed the influence of host heterogeneity and explored corrections using individual-level risk factors.

Main Results:

• Statistical network inference estimators converged to true interactions, demonstrating satisfactory performance in simulations.
• Accurate reconstruction of complex interaction networks was achieved, particularly with regularization or penalization for sample size.
• Host heterogeneity impacted performance but was successfully overcome by correcting for individual-level risk factors.

Conclusions:

• Statistical network inference is a powerful tool for detecting multi-strain pathogen interactions from population-level survey data.
• The developed methods can accurately reconstruct heterogeneous interaction networks, accounting for indirect effects.
• This approach holds significant potential for improving epidemiological studies and informing targeted disease interventions.