Adaptive dynamic ϵ-simulated annealing algorithm for tumor immunotherapy

Area of Science:

• Computational oncology
• Mathematical modeling
• Bioinformatics

Background:

• Personalized cancer therapy necessitates precise scheduling of multiple agents.
• Optimizing treatment regimens is complex due to nonlinear tumor-immune dynamics and feasibility constraints.

Purpose of the Study:

• To develop an intelligent optimization approach for complex cancer treatment models.
• To address challenges in scheduling therapeutic agents under biological constraints.

Main Methods:

• An Adaptive Dynamic ϵ-Simulated Annealing (ADϵSA) algorithm was developed.
• The algorithm integrates multi-population search, dynamic ϵ-constraint control, and boundary-aware mutation.
• ADϵSA was applied to an improved tumor immunotherapy model (ITIT) using ordinary differential equations (ODEs).

Main Results:

• ADϵSA demonstrated strong global search capability, fast convergence, and solution stability on benchmark functions.
• The algorithm identified optimal drug schedules for the ITIT model, reducing simulated tumor burden from ~1500 to below 500 cells.
• Treatment remained within physiologically acceptable limits.

Conclusions:

• ADϵSA offers advantages over traditional methods like PSO and GA for dynamic, constraint-rich ODE systems.
• This work highlights the potential of biologically informed optimization in personalized oncology.
• The study provides a computational foundation for future closed-loop, patient-specific cancer treatment strategies.