Adaptive networks: Coevolution of disease and topology

Vincent Marceau, Pierre‐André Noël, Laurent Hébert‐Dufresne, Antoine Allard, Louis J. DubéPhysical 2010 · DOI

TL;DR

This paper introduces a novel analytical framework to accurately model the simultaneous evolution of disease spread and network structure in "adaptive networks." By accounting for how individuals change their connections based on disease risk, this work provides a more realistic understanding of epidemic dynamics and their impact on network topology.

What Problem Does This Paper Address?

Traditional models of disease spread on networks often assume a static network structure, meaning the connections between individuals don't change over time. However, in reality, people can alter their behavior, including their social contacts, in response to threats like disease outbreaks. "Adaptive networks" aim to capture this dynamic interplay, where the disease influences the network structure, and the changing network structure, in turn, affects disease spread. While previous analytical approaches could identify some novel behaviors like bistability, they struggled to accurately capture the simultaneous evolution of both the disease and the network. This paper addresses this gap by developing a more sophisticated analytical method to model this coevolutionary process.

Methods

The authors build upon the adaptive SIS (Susceptible-Infectious-Susceptible) model proposed by Gross et al. Their key innovation is an "improved compartmental formalism." Instead of just tracking whether a node is susceptible or infectious, they categorize nodes based on their infectiousness state, their total degree (number of connections), and their infectious degree (number of connections to infectious individuals). This finer-grained classification allows them to derive a system of ordinary differential equations (ODEs) that describe the complete time evolution of both the disease and the network topology. They then numerically integrate these ODEs to simulate the system's behavior and compare it with Monte Carlo simulations.

Datasets

Not discussed in the paper. The authors generate synthetic networks with specific initial degree distributions (degree-regular, Poisson, and truncated power-law) for their simulations.

Key Insights and Ideas

  • Coevolution is Crucial: The interplay between disease dynamics and network topology is not just a minor detail; it fundamentally shapes epidemic outcomes and network structure.
  • Adaptive Rewiring's Dual Role: Link rewiring can both hinder disease spread by isolating infectious individuals and promote it by creating highly connected "superspreader" nodes among the susceptible population.
  • Initial Conditions Matter (for Outcome): The initial disease prevalence and network topology significantly influence whether an epidemic dies out or persists, particularly in the presence of adaptive rewiring.
  • Endemic State Universality: Once an epidemic reaches a stable endemic state, its prevalence and topology appear to be largely independent of the initial network structure, depending instead on the disease and rewiring parameters.
  • Improved Formalism's Power: The new compartmental approach accurately captures complex dynamics that simpler models miss, especially in less homogeneous networks.

Key Results

  • The developed analytical framework accurately predicts the time evolution of disease prevalence and network degree distributions, showing excellent agreement with Monte Carlo simulations across various initial network configurations (degree-regular, Poisson, power-law).
  • They demonstrate that adaptive rewiring can lead to bistable regimes (regions where either disease-free or endemic states are stable) that are not present in static networks.
  • The invasion threshold (where a disease-free state becomes unstable) for adaptive networks is shown to increase significantly with rewiring rate, particularly in homogeneous (degree-regular) networks.
  • The paper provides strong evidence that the stationary endemic state's properties (prevalence and degree distribution) are universal, depending only on the disease parameters ($\alpha, \beta, \gamma$) and the average network degree ($\langle k \rangle$), not the initial network topology.

Limitations and Future Work

  • Analytical Solution for Equilibria: The authors acknowledge that their improved compartmental formalism does not admit an analytical solution for the system's equilibria, making it impossible to mathematically prove the universality of the endemic state. This remains a conjecture.
  • Oscillatory Regimes: While oscillations were theoretically predicted in simpler models, their analytic framework currently doesn't explicitly capture them, though the authors believe it could be extended.
  • Model Realism: The authors suggest future work could incorporate more realistic features, such as community structures (cliques) and preferential rewiring (where individuals might avoid highly connected people). This could lead to more nuanced understanding of disease control strategies.

Summary updated

Abstract

Adaptive networks have been recently introduced in the context of disease propagation on complex networks. They account for the mutual interaction between the network topology and the states of the nodes. Until now, existing models have been analyzed using low complexity analytical formalisms, revealing nevertheless some novel dynamical features. However, current methods have failed to reproduce with accuracy the simultaneous time evolution of the disease and the underlying network topology. In the framework of the adaptive susceptible-infectious-susceptible (SIS) model of Gross [Phys. Rev. Lett. 96, 208701 (2006)]10.1103/PhysRevLett.96.208701, we introduce an improved compartmental formalism able to handle this coevolutionary task successfully. With this approach, we analyze the interplay and outcomes of both dynamical elements, process and structure, on adaptive networks featuring different degree distributions at the initial stage.
BibTeX
@inproceedings{marceau2010adaptive,
  author = {Vincent Marceau and Pierre‐André Noël and Laurent Hébert‐Dufresne and Antoine Allard and Louis J. Dubé},
  title = {{Adaptive networks: Coevolution of disease and topology}},
  booktitle = {Physical},
  year = {2010},
  doi = {10.1103/physreve.82.036116},
}
Figure