Evolutionary Games with Affine Fitness Functions: Applications to Cancer

Moritz Gerstung, Hani Nakhoul, Niko BeerenwinkelDyn. Games Appl 2011 · DOI

TL;DR

This paper extends evolutionary game theory to model cancer progression by introducing "affine fitness functions," which account for both interactions between cell types and intrinsic cell-specific advantages. They demonstrate how these affine terms can enable cooperation in scenarios where it would otherwise be impossible and apply this framework to model tumor-normal cell interactions, identifying strategies that promote tumor growth.

What Problem Does This Paper Address?

Evolutionary game theory is a powerful tool for understanding how populations change over time based on the fitness of different strategies. Traditionally, fitness in these models is often a linear function of the expected payoff from interactions. However, biological systems, particularly cancer, are more complex. Cancer cells acquire mutations that can confer intrinsic advantages independent of interactions, and their interactions with normal cells (stroma) are crucial for tumor growth. Existing models often struggle to capture these dual influences. This paper addresses this gap by introducing "affine fitness functions," which combine a linear payoff term from interactions with a constant, cell-type-specific fitness contribution. This allows for a more nuanced representation of evolutionary dynamics, particularly in the context of cancer, where intrinsic proliferation advantages and complex interactions with the tumor microenvironment are key drivers of progression.

Methods

The core methodological contribution is the introduction and analysis of affine fitness functions in evolutionary game theory. This is achieved by:

  • Extending the Replicator Equation: The standard replicator equation, which describes the dynamics of strategy frequencies in infinite populations, is modified to incorporate the affine fitness function. A key theoretical result (Theorem 1) shows that this inhomogeneous replicator equation is equivalent to a homogeneous replicator equation with a transformed payoff matrix. This transformation is crucial as it can lead to substantially different dynamics than the original game.
  • Analyzing Two-Player Games: The authors analyze the stability of equilibria in two-player games (e.g., Prisoner's Dilemma) with affine fitness. They derive conditions under which previously impossible outcomes, like the evolution of cooperation, become possible due to the constant fitness terms.
  • Stochastic Dynamics in Finite Populations: The paper extends the analysis to finite populations using the Moran process. They investigate how affine fitness terms affect criteria for evolutionary stability, advantageousness, and risk dominance, deriving specific conditions for these in both small and large populations.
  • Modeling Tumor-Normal Interactions: The framework is applied to model interactions between tumor cells and normal cells. This involves extending the game to three players (normal cell type and two tumor cell types) and analyzing the conditions under which different tumor strategies can coexist with or dominate normal cells.

Datasets

Not discussed in the paper. The paper focuses on theoretical modeling and analysis, not empirical data fitting or training on specific datasets.

Key Insights and Ideas

  • Affine Fitness Enables Cooperation: The addition of a constant fitness term can transform games like the Prisoner's Dilemma, making cooperation evolutionarily stable even when it would be impossible with linear fitness alone. This highlights how intrinsic advantages can overcome disadvantages in direct interactions.
  • Transformed Game Dynamics: While an affine fitness function can be mapped to an equivalent homogeneous game, the dynamics of this transformed game can be significantly different from the original. This means simply transforming the payoff matrix doesn't fully capture the evolutionary consequences of affine fitness.
  • Tumor Strategies and Coexistence: In the context of cancer, affine fitness allows for modeling tumor cells that both exploit normal cells and possess intrinsic proliferation advantages. The paper identifies specific tumor strategies that are more successful in establishing themselves within normal tissue, often by balancing exploitation with recruitment of normal cells.
  • Constant Fitness vs. Interaction Payoff: The analysis of tumor-normal interactions reveals a critical trade-off. The success of a tumor strategy depends on the balance between its constant fitness advantage and its payoff from interacting with normal cells, as well as the cost of these interactions.

Key Results

  • Prisoner's Dilemma: For a Prisoner's Dilemma game, cooperation can evolve if the constant fitness advantage for cooperators is greater than the cost of cooperation minus the benefit received by defectors, scaled by the population size.
  • Tumor-Normal Interactions (Two-Player): For a pairwise game between a tumor and normal cell, a stable equilibrium exists if the constant fitness advantage of the tumor is within a specific range determined by the payoff differences.
  • Tumor-Normal Interactions (Three-Player): In a game involving normal cells and two tumor strategies, the dynamics are complex and depend on the parameter 'b' (payoff to normal cells) and the constant fitness advantages (σ) of the tumor strategies. The paper identifies regimes where normal cells go extinct, where all cell types coexist, or where one tumor strategy dominates another. Strategy T3, with the largest constant fitness advantage, was found to be most successful in outcompeting other tumor strategies for most parameter values.

Limitations and Future Work

  • Parameter Estimation: The authors acknowledge that most parameters in their models cannot be directly obtained from experiments, making direct quantitative prediction challenging.
  • Simplified Interactions: The models often simplify interactions, particularly in the three-player game where tumor-tumor interactions are initially assumed to be absent.
  • Future Work: The paper suggests incorporating mutations to model the emergence of new cell types, which is a crucial aspect of cancer evolution. Extending the analysis to include spatial dynamics and more complex interaction networks are also promising avenues. The authors also note that while their model can explain how normal cells might be attracted to tumors, achieving therapeutic success by eliminating tumor cells requires a careful balance of reducing interaction benefits and increasing intrinsic costs for tumor cells.

Summary updated

BibTeX
@article{gerstung2011evolutionary,
  author = {Moritz Gerstung and Hani Nakhoul and Niko Beerenwinkel},
  title = {{Evolutionary Games with Affine Fitness Functions: Applications to Cancer}},
  journal = {Dyn. Games Appl.},
  year = {2011},
  doi = {10.1007/s13235-011-0029-0},
}
Figure