Stochastic Evolutionary Control in Heterogeneous Populations

Area of Science:

• Evolutionary biology
• Computational biology
• Mathematical modeling

Background:

• Therapeutic resistance is a major challenge in treating cancer and infectious diseases.
• Existing methods often rely on the strong-selection, weak-mutation assumption, limiting their applicability to homogeneous populations.
• Understanding the evolutionary dynamics of heterogeneous populations is crucial for developing effective resistance-mitigation strategies.

Purpose of the Study:

• To introduce SHEPHERD (Stochastic Heterogeneity-informed Evolutionary Policy Hampering the Expansion of Resistance to Drugs), a novel framework for designing adaptive drug policies.
• To move beyond the strong-selection, weak-mutation assumption by incorporating stochastic heterogeneity in evolutionary modeling.
• To reduce the fitness of evolving diseases and mitigate therapeutic resistance.

Main Methods:

• Integration of Wright-Fisher (WF) population genetics modeling with Markov Decision Processes (MDPs).
• Development of a computational framework (SHEPHERD) to capture full stochastic dynamics of genetically heterogeneous populations.
• Simulation using synthetic multi-drug fitness landscapes with varying numbers of genotypes.

Main Results:

• The optimized SHEPHERD protocol significantly reduced long-term mean disease fitness compared to single-drug or periodic two-drug regimens in silico.
• Demonstrated robustness of SHEPHERD strategies across different levels of genotypic frequency discretization and temporal resolution.
• Highlighted strong dependence of strategy effectiveness on the timing of drug updates.

Conclusions:

• SHEPHERD provides a foundation for applying MDP optimization to control a broad class of evolving populations beyond the strong-selection, weak-mutation regime.
• The approach offers a promising strategy for mitigating therapeutic resistance in complex, heterogeneous disease systems.
• Acknowledged the need for empirical multi-genotype, multi-drug measurements to further validate and refine the models.