Targeted cancer therapy by nonlinear data-enabled predictive control (DeePC): A case study on lewis lung carcinoma

This paper investigates the application of the Data-Enabled Predictive Control (DeePC) framework for managing and controlling the growth of Lewis lung carcinoma (LLC). Unlike conventional approaches that require detailed knowledge of the cellular or molecular mechanisms of cancer, DeePC relies exclusively on input–output data collected during an initial learning phase. In this phase, pseudo-random binary sequences (PRBS) of Angiostatin and Endostatin are administered at safe dose limits of 20mg/kg, and tumor volume measurements are recorded daily over a period of 62 to 116 days. The predictive capabilities of DeePC enable periodic control and measurement updates (typically every 10 days), which effectively manage tumor growth under varying levels of uncertainty. The study evaluates the nonlinear tumor dynamics under scenarios of 1% to 31% uncertainty in the state equations and output measurements, demonstrating the robustness of DeePC to noise and disturbances. Simulation results show that the framework can suppress tumor volume to negligible levels using low average doses, though performance depends on the selected DeePC parameters and the level of measurement and process noise. These findings underscore the potential of DeePC as a safe, efficient, and fully data-driven approach to targeted cancer therapy.