Deep reinforcement learning-based control of multi-drug cancer models
Abstract
Cancer remains a critical threat to human life and health, with the emergence of drug resistance standing as the greatest challenge in cancer treatment. To mitigate the adverse impacts of cancer drug resistance, current research efforts have focused on the development of single-drug and multi-drug models, as well as the formulation of corresponding therapeutic strategies. Owing to the relative simplicity of the single-drug framework, the development of models and the optimization of treatment strategies under this framework have reached a mature stage. In contrast, the multi-drug framework, characterized by its inherent complexity, still lacks fully developed models and optimized strategies. In this study, we established a dual-drug cancer treatment model and employed deep reinforcement learning algorithms, widely studied and applied in the single-drug framework, for control optimization. Additionally, we analyzed the effects of complex environmental factors, such as discrete drug dosing, drug synergism, and unidirectional collateral sensitivity, on the optimization of cancer treatment strategies. The results demonstrated that deep reinforcement learning algorithms still yield excellent optimization outcomes under the dual-drug framework. We found that complex environmental factors in the multi-drug framework exert varying degrees of influence on all cancer treatment optimization strategies. In future work, it is necessary to identify and screen out the complex environmental factors with the greatest impact and formulate specific optimization strategies to delay the emergence of drug resistance to the maximum extent.