Adaptive Dynamic ε-Simulated Annealing Algorithm for Tumor Immunotherapy

Xiaoyan Sun¹Ying Jiang^2,3*

• ¹The People's Hospital of Liaoning Province, Shenyang, Liaoning Province, China
• ²Department of Gynecology, Liaoning Cancer Hospital, China Medical University, Shenyang, China
• ³Liaoning Cancer Hospital, China Medical University, Shenyang, Liaoning Province, China

In this study, we present an Adaptive Dynamic ε-Simulated Annealing (ADεSA) algorithm tailored for the optimization of tumor immunotherapy strategies. This work is entirely computational, based on a nonlinear ordinary differential equation (ODE) model that captures tumor-immune-drug dynamics, with parameters adapted from previously validated biological studies. Unlike traditional metaheuristics such as PSO or GA, ADεSA incorporates a multi-population structure, a dynamic ε-constraint adaptation mechanism, and boundary-aware mutation control, making it particularly suited to constrained, trajectory-sensitive, and biologically grounded optimization problems. The proposed method is integrated with an improved tumor immunotherapy model (ITIT), enabling the personalized scheduling of chemotherapeutic and immunotherapeutic drug dosages. The optimization simultaneously accounts for treatment intensity, toxicity control, and immune system sustainability, reflecting clinical considerations such as dosage tapering and immune modulation. Simulation results on 12 benchmark functions demonstrate that ADεSA achieves faster convergence, greater stability, and stronger global search capability than standard approaches. When applied to the ITIT model, the algorithm significantly reduces simulated tumor cell populations (from ~1500 to ~500) while maintaining drug dosing within feasible clinical ranges. This work highlights the potential of intelligent optimization frameworks to support personalized, adaptive oncology, and provides a computational foundation for real-time treatment planning in future applications.