About

Contact

For questions about epymorph or the overall EpiMoRPH project, please email:

Motivation for epymorph

Metapopulation models are one of the most powerful tools available to study and understand infectious disease dynamics, forecast pathogen spread during outbreaks, and guide the timing and nature of public health interventions in managing outbreaks. Outbreaks of pathogens including influenza virus, Ebola virus, West Nile virus, Mpox, and SARS-CoV-2 have emphasized the need for a responsive modeling community and robust modeling infrastructure, capable of quickly developing models that accurately characterize and forecast the dynamics of pathogen outbreaks across the globe.

The Problem: Models as walled gardens

Modelers have created increasingly sophisticated models and demonstrated their efficacy in capturing and predicting disease outbreak dynamics. Unfortunately, modeling code infrastructures are often custom constructions, painstakingly designed, coded, and tailored by a particular modeler for a particular location or application scenario. This introduces a number of significant limitations, particularly in the face of rapidly evolving outbreak scenarios, including:

  • Models take days or weeks of custom coding to construct and fine-tune.
  • Specifics of particular locales or pathogens are often hard-wired into the model code in many levels or places.
  • Models may hard-wire in various assumptions and dynamics within the modeling code, making it challenging to articulate and understand these vital factors.
  • The modeling pipelines themselves, are typically private, not accessible to outside evaluators of the model’s performance.

Together, these issues make it challenging to understand model assumptions and dynamics, compare why one model may produce different dynamics than another model, and evaluate a model’s applicability to other locales or scenarios.

More generally, we do not know of any wholistic modeling software frameworks that capture and support the full model lifecycle, particularly for spatial disease models. We define the modeling lifecycle here as extending from specification of compartment model equations, to specification of host movement dynamics, to gathering of data relevant to geographic attributes (e.g., node locations, population sizes, distances between nodes, environmental characteristics across nodes, etc.), to the simulation of models, to estimation of dynamic model parameters against real surveillance data.

The Solution: a modular, universal metapopulation modeling framework

The epymorph package is the product of the EpiMoRPH (Epidemiological Modeling Resources for Public Health) project, and aims to address the shortcomings outlined above by:

  1. Providing a uniform, streamlined framework for building, testing, fitting, and evaluating metapopulation models for infectious pathogens;
  2. that is easily accessible to beginning modelers, while also sophisticated enough to allow rapid design and execution of complex modeling experiments by highly experienced modelers, and;
  3. streamlines and improves modeling quality by dramatically decreasing model construction times, increasing model transparency, automatically retrieving key modeling data from public databases, automatically fitting models to observed data, and ensuring fast, easy portability and performance evaluation of models across temporal and geographic scenarios.

More broadly, the EpiMorph project aims to provide a collaborative virtual space can (selectively) share various modeling components, rapidly construct and test new models by combining their own and shared components, and rapidly evaluate the performance of models against evolving public health data.

Acknowledgments

The development of epymorph was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award number R01AI168144 to PI Joseph Mihaljevic, and by the Southwest Health Equity Research Collaborative at Northern Arizona University (U54MD012388), which was sponsored by the National Institute on Minority Health and Health Disparities (NIMHD).