Tracking the intractable: The role of dynamic network models in complex systems

What do infectious disease outbreaks and traffic congestion have in common? Both are examples of complex systems, which shape our world in ways that are often intractable. Computational models of such systems can enable us to understand their processes better, and even make predictions, like weather forecasts. Capturing the intricacies of these systems in computational models is a long-standing challenge in research, but fine-grained, dynamic approaches offer a solution. Utrecht University researcher Deb Panja is specialised in such models, and together with his collaborators, he aims to develop better understanding of complex systems, potentially improving the ways in which we can try to control them.

From the spread of infectious diseases to the emergence of traffic jams, many challenges in society are ‘wicked’ or complex systems. These are systems in which small, local interactions can lead to large-scale nontrivial outcomes. Traditional computational models often struggle to bridge such a large gap in scales.

Deb Panja, who specialises in developing fine-grained, dynamic network models, is pioneering approaches that account for the heterogeneity of real-world interactions in space and time. Two recently published studies co-authored by Panja illustrate how this approach improves our understanding of complex systems.

The power of dynamic network models

The need for models that can capture micro-level variability is clear: human behaviour, and other interconnected elements within a network, are stochastic, unpredictable, contexts vary widely, and interactions are constantly changing. By integrating detailed system actors, spatial, and temporal data, researchers can uncover patterns that static network models miss. This is precisely what Panja and his collaborators achieve in their works, offering a more nuanced, realistic and scalable way to model complex systems.

Accounting for heterogeneity and real-time interactions leads to more accurate representations of reality and therefore better-informed interventions

Mapping the spread of pathogens

In the first study, published in Communications Medicine, the researchers developed a large-scale, actor-based model to map the spread of respiratory pathogens like influenza and SARS-CoV-2. By incorporating detailed data from the Dutch population registry, they showed how the geographic and demographic profiles of initial cases influence transmission. The model enables rigorous assessment of intervention strategies incorporating behavioural adaptations, which highlights their potential to modulate epidemic outcomes.

Superspreading events and transport delays

The second study, in Nature Communications, takes this idea further by introducing event-based spatiotemporal networks – dynamic networks anchored in space and time – to model emergent phenomena. The researchers applied this framework to track a novel respiratory pathogen in the Netherlands, revealing transmission routes, infection patterns, and superspreading events. They also demonstrated its versatility by modelling the propagation of delays in Switzerland’s public transport system. The researchers argue that their approach can be used to model emergent phenomena in a broad range of fields, such as developmental biology and community ecology.

The article was featured on the Editors’ Highlights webpage, which showcases the best papers recently published in an area.

Future work

These studies show that fine-grained, dynamic models are not just theoretical improvements, they are a practical necessity. Whether in epidemiology, urban planning, or other fields, accounting for heterogeneity and real-time interactions leads to more accurate representations of reality and therefore better-informed interventions. As the researchers note, future work will focus on refining these models to balance detail with scalability, making them even more powerful tools for simplifying the complex systems that shape our world.