Gaussian Process Emulation for Exploring Complex Infectious Disease Models.

Area of Science:

• Epidemiology
• Computational Biology
• Mathematical Modeling

Background:

• Complex epidemiological models offer biological realism but face computational challenges due to numerous parameters.
• Individual-based models (IBMs) for disease dynamics are powerful but computationally intensive, limiting parameter space exploration.

Purpose of the Study:

• To investigate Gaussian Process (GP) emulation as a method to overcome computational limitations in complex IBMs.
• To develop and validate GP surrogate models for predicting dengue transmission dynamics and outcomes.

Main Methods:

• Developed an individual-based model (IBM) for dengue transmission incorporating social structure, seasonality, and human movement.
• Trained three GP surrogate models on key outcomes: outbreak probability, maximum incidence, and epidemic duration.
• Utilized a dataset of over 1,000 dengue epidemics in Colombia (12 years) for calibration and validation.

Main Results:

• GP emulation enabled rapid prediction of epidemiological outcomes across an eight-dimensional parameter space.
• Identified average infectivity and human mobility as primary drivers of dengue outbreak metrics.
• Seasonal timing of initial infection was found to influence epidemic course.

Conclusions:

• GP emulation significantly enhances the feasibility of using complex, realistic IBMs in epidemiological research.
• The calibrated GP model successfully identified high-risk areas for targeted public health interventions in Colombia.
• This approach holds promise for improving disease control strategies for diseases like dengue.