Network topological determinants of pathogen spread

María Pérez‐Ortiz, Petru Manescu, Fabio Caccioli, Delmiro Fernández-Reyes, Parashkev Nachev, John Shawe‐TaylorScientific 2022 · DOI

Abstract

How do we best constrain social interactions to decrease transmission of communicable diseases? Indiscriminate suppression is unsustainable long term and presupposes that all interactions carry equal importance. Instead, transmission within a social network has been shown to be determined by its topology. In this paper, we deploy simulations to understand and quantify the impact on disease transmission of a set of topological network features, building a dataset of 9000 interaction graphs using generators of different types of synthetic social networks. Independently of the topology of the network, we maintain constant the total volume of social interactions in our simulations, to show how even with the same social contact some network structures are more or less resilient to the spread. We find a suitable intervention to be specific suppression of unfamiliar and casual interactions that contribute to the network's global efficiency. This is, pathogen spread is significantly reduced by limiting specific kinds of contact rather than their global number. Our numerical studies might inspire further investigation in connection to public health, as an integrative framework to craft and evaluate social interventions in communicable diseases with different social graphs or as a highlight of network metrics that should be captured in social studies.

Introduction

During public health crises like the COVID-19 pandemic, indiscriminate suppression of social interactions—commonly known as lockdowns—often serves as a primary tool for disease control. However, such measures are economically and socially unsustainable. While it is well-established that the "shape" or topology of social networks influences transmission, most theoretical models analyze metrics like average degree in isolation. In reality, network features are deeply entangled; changing one often inadvertently alters others, such as clustering or path length. This study aims to untangle these determinants to identify specific topological features that drive pathogen spread, moving toward prescriptive models for targeted interventions.

Methods

We employed a large-scale simulation framework using a stochastic Susceptible-Exposed-Infectious-Recovered (SEIR) model. A critical component of our methodology was maintaining a constant total volume of social interactions across all simulations. This experimental control allowed us to isolate the impact of network structure from the sheer quantity of contact.

We generated a dataset of 9,000 interaction graphs across four families: Erdos-Renyi (random), Scale-free (heavy-tailed), Small-world, and Community-based networks. Networks varied in size ($N=500$ to $2,000$) and connectivity (mean degrees of 4, 10, and 20). Pathogen parameters were initialized based on SARS-CoV-2 estimates. Interactions were categorized as "local" (static, repeated contacts) or "global" (dynamic, casual contacts). We analyzed 30 topological metrics to determine their predictive power regarding the cumulative percentage of infected individuals.

Results

The Primacy of Global Efficiency

Our analysis revealed that global efficiency—a measure of how easily a virus can exchange across the entire network—is the strongest predictor of epidemic severity. A linear regression model incorporating all 30 topological metrics achieved an adjusted $R^2$ of 0.934, indicating that topology almost entirely explains the variability in spread. Notably, global efficiency alone ($R^2 = 0.747$) outperformed the combined predictive power of all local degree-based metrics ($R^2 = 0.710$). Spectral metrics, such as algebraic connectivity, were also identified as key indicators of a network's vulnerability to rapid diffusion.

Impact of Global Interactions

The addition of "global" interactions significantly increases the total percentage of the population infected, even when the total volume of interactions remains constant. This effect is particularly pronounced in networks with lower average degrees, where the introduction of casual contacts dramatically shifts the infection outcome.

Impact of global interactions on infection spread across different average degrees.

Topological Resilience and Sensitivity

Different network architectures exhibit varying levels of resilience. For instance, while scale-free networks experience rapid initial spread due to high-degree hubs, these hubs often become immune early, acting as firebreaks. Conversely, small-world networks are highly sensitive to global interactions. These casual contacts act as shortcuts that bypass local clusters, causing the epidemic to explode.

Distribution of infected individuals across different network topologies with and without global interactions.

Discussion

Our findings suggest that the total volume of social interaction is a less critical determinant of pathogen spread than the specific topology of those interactions. Specifically, suppressing unfamiliar and casual interactions—which contribute most to global efficiency—is a highly effective intervention. This approach reduces spread without requiring the drastic reduction in total social contact associated with indiscriminate lockdowns.

Future work should focus on scaling these simulations to larger, real-world social networks and integrating them with empirical mobility data. While the SEIR model used here assumes total immunity post-infection, future iterations could account for waning immunity and re-infection to better reflect the dynamics of evolving pathogens.

Summary updated

Abstract

How do we best constrain social interactions to decrease transmission of communicable diseases? Indiscriminate suppression is unsustainable long term and presupposes that all interactions carry equal importance. Instead, transmission within a social network has been shown to be determined by its topology. In this paper, we deploy simulations to understand and quantify the impact on disease transmission of a set of topological network features, building a dataset of 9000 interaction graphs using generators of different types of synthetic social networks. Independently of the topology of the network, we maintain constant the total volume of social interactions in our simulations, to show how even with the same social contact some network structures are more or less resilient to the spread. We find a suitable intervention to be specific suppression of unfamiliar and casual interactions that contribute to the network's global efficiency. This is, pathogen spread is significantly reduced by limiting specific kinds of contact rather than their global number. Our numerical studies might inspire further investigation in connection to public health, as an integrative framework to craft and evaluate social interventions in communicable diseases with different social graphs or as a highlight of network metrics that should be captured in social studies.
BibTeX
@inproceedings{perezortiz2022network,
  author = {María Pérez‐Ortiz and Petru Manescu and Fabio Caccioli and Delmiro Fernández-Reyes and Parashkev Nachev and John Shawe‐Taylor},
  title = {{Network topological determinants of pathogen spread}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41598-022-11786-5},
}
Figure