Dr. Anna Maria Langmüller, MSCA postdoctoral fellow, Cornell University, US
The role of spatial and environmental heterogeneity in disease dynamics
The use of ordinary differential equations to analyze infectious disease dynamics is a widely accepted practice in epidemiology. However, standard models such as the classical SIR model typically assume homogeneously mixing populations, which fails to account for potential spatial heterogeneity in the prevalence and spread of a disease. Reaction-diffusion models can offer a solution to this limitation by accounting for both temporal and spatial variations in the frequency of susceptible and infectious individuals across a continuous landscape. In the first part of my talk, we will explore the conditions under which such spatial structure must be explicitly considered to accurately predict disease spread through a population, and when a model assuming homogeneous mixing remains adequate. We validate our predictions using individual-based simulations implemented in the versatile evolutionary simulation framework SLiM4 and identify potential design considerations and limitations for simulation models of disease dynamics. Spatial structure is only one of potentially many factors that can shape disease dynamics. Individual-based simulation frameworks have proven to be instrumental when studying real-world epidemics, because they provide the necessary flexibility to build disease models of arbitrary complexity. However, detailed sensitivity analyses of these often parameter-rich models remain challenging, because the necessary model recalibration is computationally expensive. In the second part of my talk, we will explore Gaussian processes as a powerful statistical framework that could significantly advance sensitivity analyses for individual-based models, allowing us to assess the relative input importance of different model parameters in unprecedented detail. This could not only further deepen our understanding of real-world disease dynamics, but also allow us to evaluate the potential of proposed intervention strategies.