Multiscale mathematical model-informed reinforcement learning optimizes combination treatment scheduling in...

Area of Science:

• Computational Biology
• Mathematical Oncology
• Artificial Intelligence in Medicine

Background:

• Tumor-microenvironment interactions significantly influence cancer growth and drug resistance, posing challenges for treatment optimization.
• Mathematical models are crucial for understanding these dynamics and developing effective therapeutic strategies, particularly in complex cancers like glioblastoma.

Purpose of the Study:

• To develop and validate a multiscale mathematical model-informed reinforcement learning (M4RL) framework for simulating tumor-microenvironment dynamics.
• To optimize dynamic scheduling of combination drug therapies for glioblastoma, focusing on immunotherapy resistance.
• To identify the most effective drug combination regimen and treatment schedule.

Main Methods:

• Development of a multiscale agent-based model (MSABM) to simulate interactions between tumor cells (TCs) and tumor-associated macrophages (TAMs).
• Learning a surrogate model from the MSABM using physics-informed neural networks and Fokker-Planck equations.
• Employing a surrogate model-based reinforcement learning approach with an actor-critic algorithm for treatment scheduling optimization.

Main Results:

• The M4RL framework successfully simulated dynamic tumor-microenvironment interactions.
• An optimal dynamic combination regimen of CSF1R inhibitor and IGF1R inhibitor was identified for glioblastoma treatment.
• The predicted optimal regimen was validated using spatial transcriptomic data, demonstrating its efficacy.

Conclusions:

• The M4RL framework provides a powerful computational tool for dissecting complex tumor-microenvironment interactions.
• This approach enables the optimization of dynamic drug combination scheduling, offering a path towards personalized cancer therapy.
• The study highlights the potential of integrating mathematical modeling and machine learning for advancing cancer treatment strategies.