Optimal Control of Heterogeneous Mutating Viruses

Elena A. Gubar, Vladislav Taynitskiy, Quanyan ZhuGames 2018 · DOI

TL;DR

This paper introduces a novel framework that couples epidemic modeling (SIR model) with evolutionary game theory to study the spread of mutating viruses. By formulating an optimal control problem, the authors derive strategies for medical treatment and quarantine to minimize epidemic costs, demonstrating that these strategies have a threshold structure dependent on cost functions.

What Problem Does This Paper Address?

Traditional epidemic models, like the SIR model, often simplify viruses as static entities. However, real-world pathogens, such as influenza, constantly mutate, leading to new strains with potentially different virulence and transmissibility. This mutation process can significantly impact epidemic dynamics, making it crucial to account for both the spread of the virus within a population and its evolution. The challenge lies in developing models that capture this interplay and devising effective control strategies that adapt to evolving viral threats, considering factors like treatment costs and quarantine effectiveness.

Methods

The authors propose a coupled system that integrates the standard SIR model for human population dynamics with replicator dynamics from evolutionary game theory to model virus mutations.

  • SIR Model Extension: The SIR model is extended to account for two distinct virus strains (V₁ and V₂). The population is divided into susceptible (S), infected by V₁ (I₁), infected by V₂ (I₂), and recovered (R) compartments.
  • Virus Evolution: Replicator dynamics are used to model the competition and mutation between the two virus strains. This involves defining fitness functions for each virus type based on their interaction with hosts and their energy costs. The model captures how the proportion of each virus strain changes over time based on their relative success.
  • Optimal Control Formulation: An optimal control problem is formulated to minimize the total cost over a given time interval. This cost includes infected costs, benefits from recovered individuals, and costs associated with medical treatments (quarantine/intensive care). The control variables are the fractions of infected individuals undergoing treatment or quarantine for each virus type (u₁ and u₂).
  • Pontryagin's Maximum Principle: This principle is employed to derive the structure of the optimal control strategies. It involves defining a Hamiltonian, constructing an adjoint system, and analyzing the conditions for optimality.

Datasets

Not discussed in the paper. The study relies on simulated data generated through numerical experiments.

Key Insights and Ideas

  • Virus Mutation Matters: The study highlights that ignoring virus mutation can lead to suboptimal control strategies and higher overall epidemic costs.
  • Coupled Dynamics are Crucial: Effectively managing epidemics requires understanding the interplay between virus evolution and population dynamics.
  • Threshold Control Strategies: The optimal control strategies for medical treatment and quarantine exhibit a threshold structure. This means that control is either fully applied or not applied at all, switching at specific time points determined by the cost functions and the state of the epidemic.
  • Cost Function Influence: The nature of the cost functions (concave vs. convex) for medical treatments significantly influences the structure of the optimal control. Convex costs lead to a more continuous control strategy, while concave costs result in a "bang-bang" control (either fully on or fully off).
  • Nash Equilibrium for Virus Evolution: The evolutionary game model predicts Nash equilibria for the virus population, indicating the stable proportions of different virus strains that can emerge.

Key Results

  • Reduced Costs: Numerical simulations demonstrate that applying optimal control strategies significantly reduces aggregated epidemic costs compared to uncontrolled scenarios. For instance, in one experiment, costs dropped from 1004.43 to 268.35.
  • Effective Epidemic Termination: Optimal control strategies were shown to effectively reduce the infected population to zero by the end of the epidemic period, leading to a high proportion of recovered individuals.
  • Threshold Behavior Observed: The simulations visually confirm the threshold structure of the optimal control strategies, showing periods of intense intervention followed by periods of no intervention.
  • Impact of Virus Strength: The study shows how differences in virus virulence (represented by infection rates) influence the epidemic's progression and the effectiveness of control measures.

Limitations and Future Work

  • Limited Virus Types: The current model considers only two virus types. Future work could extend this to multiple virus strains.
  • Latent Individuals: The study acknowledges that it does not explicitly model latent individuals (infected but asymptomatic), a subgroup that can significantly influence epidemic dynamics. Incorporating this group would enrich future research.
  • Imitative and Best-Response Dynamics: The authors suggest exploring other evolutionary dynamics, such as imitative dynamics and best-response dynamics, to model virus mutations.
  • Network Structures: The current model assumes a well-mixed population. Future research could investigate the impact of network structures on epidemic spread and control.

Summary updated

BibTeX
@article{gubar2018optimal,
  author = {Elena A. Gubar and Vladislav Taynitskiy and Quanyan Zhu},
  title = {{Optimal Control of Heterogeneous Mutating Viruses}},
  journal = {Games},
  year = {2018},
  doi = {10.3390/g9040103},
}
Figure