SourceR

SourceR attributes human cases of food-borne zoonotic diseases to animal or environmental sources using strain-typed surveillance data and Bayesian non-parametric models to inform quantitative source attribution.

Key Features:

• Bayesian non-parametric modeling (Dirichlet processes): Uses Dirichlet processes to model and cluster pathogen strain types without fixed parametric assumptions.
• Strain-typed surveillance data integration: Integrates strain-typed data from human cases and source samples while accounting for differences in survivability, pathogenicity, virulence, and source-specific transmission roles.
• Force of infection quantification: Estimates the force of infection contributed by each potential source to quantify relative source importance.
• Source and type effects: Incorporates explicit source-level and strain-type effects into the attribution model to capture heterogeneous contributions.
• Strain clustering to detect high-virulence types: Identifies clusters of multilocus sequence types or strain groups associated with elevated human virulence.

Scientific Applications:

• Food safety interventions: Identifies and prioritizes animal or food sources contributing to human cases to inform targeted interventions reducing morbidity and productivity losses.
• Epidemiological research: Clusters pathogen strains by epidemiological behavior to detect strain types associated with high virulence and to study disease dynamics.
• Case attribution — Campylobacter jejuni: Applied to Campylobacter jejuni isolate data from New Zealand (2005–2008) to attribute cases to a specific poultry source and to detect a cluster of nine multilocus sequence types with unusually high human virulence.

Methodology:

Applies a Bayesian attribution framework with Dirichlet process-based non-parametric modeling to cluster pathogen strain types, incorporates source and type effects, and estimates force of infection for each source.