Reinforcement learning improves chemotherapy dosing under tumor diversity
Adaptive Chemotherapy Control under Tumor Heterogeneity via Reinforcement Learning
Machine Learning
Summary
Chemotherapy treatment is hard to get right because tumors can be very different and some cells resist drugs. The authors created smart computer programs that learn how to adjust chemotherapy doses in real time based on how tumor cells respond. They tested two learning methods and found that one reduced tumors more on average while the other gave more consistent dose amounts. These methods work well in computer simulations but assume we can see all tumor cells clearly, which is harder in real life.
What this means in practice
- •For oncology treatment planners: Use reinforcement learning to design adaptive chemotherapy schedules tailored to patient tumor differences for improved tumor reduction.
- •For pharmaceutical data scientists: Develop simulation-driven dose optimization models to evaluate drug efficacy under tumor variability conditions.
Tested on simulated data.
Authors
Bereket Sitotaw Kidane, Md Samiul Haque Motayed, Shuo Wang
Abstract
Designing effective chemotherapy regimens is hindered by tumor heterogeneity and drug resistance, which complicate the deployment of patient-specific model-based optimal control across diverse populations. We develop and compare closed-loop deep reinforcement learning (DRL) dosing policies with continuous (TD3) and discrete (DQN) action spaces trained on a high-dimensional heterogeneous tumor model. The DRL policies are benchmarked against a Pontryagin's Maximum Principle (PMP)-derived open-loop benchmark. We assess generalization under parametric heterogeneity using a 100-patient virtual cohort with plus or minus 10 percent uniform perturbations in growth and drug-sensitivity parameters. Across this cohort, TD3 achieves higher average tumor reduction, while DQN yields tighter inter-patient dosing consistency, revealing a clear efficacy-consistency trade-off in this study. Our simulations assume full observation of all tumor subpopulations; translation to sparse and noisy clinical measurements will require partial-observability formulations and/or state estimation. Overall, the results show that simulation-trained DRL can learn state-dependent feedback dosing policies that complement open-loop optimal control benchmarks.