An implementation framework to improve the transparency and reproducibility of computational models of infectious diseases

Understanding ecological systems using knowledge graphs: an application to highly pathogenic avian influenza

Motivation

Ecological systems are complex. Representing heterogeneous knowledge about ecological systems is a pervasive challenge because data are generated from many subdisciplines, exist in disparate sources, and only capture a subset of interactions underpinning system dynamics. Knowledge graphs (KGs) have been successfully applied to organize heterogeneous data and to predict new linkages in complex systems. Though not previously applied broadly in ecology, KGs have much to offer in an era when system dynamics are responding to rapid changes across multiple scales.

Results

We developed a KG to demonstrate the method's utility for ecological problems focused on highly pathogenic avian influenza (HPAI), a highly transmissible virus with a broad host range, wide geographic distribution, and rapid evolution with pandemic potential. We describe the development of a graph to include data related to HPAI including pathogen-host associations, species distributions, and population demographics, using a semantic ontology that defines relationships within and between datasets. We use the graph to perform a set of proof-of-concept analyses validating the method and identifying patterns of HPAI ecology. We underscore the generalizable value of KGs to ecology including ability to reveal previously known relationships and testable hypotheses in support of a deeper mechanistic understanding of ecological systems.

Availability and implementation

The data and code are available under the MIT License on GitHub at https://github.com/cghss-data-lab/uga-pipp.

Reproducibility of COVID-era infectious disease models

Infectious disease modelling has been prominent throughout the COVID-19 pandemic, helping to understand the virus' transmission dynamics and inform response policies. Given their potential importance and translational impact, we evaluated the computational reproducibility of infectious disease modelling articles from the COVID era. We found that four out of 100 randomly sampled studies released between January 2020 and August 2022 could be completely computationally reproduced using the resources provided (e.g., code, data, instructions) whilst a further eight were partially reproducible. For the 100 most highly cited articles from the same period we found that 11 were completely reproducible with a further 22 partially reproducible. Reflecting on our experience, we discuss common issues affecting computational reproducibility and how these might be addressed.