This white paper introduces a conceptual reinforcement learning framework for adaptive cancer therapy optimization, integrating genomic and phenotypic data within a unified computational model. The proposed approach formulates cancer treatment as a sequential decision-making problem under uncertainty, leveraging Deep Reinforcement Learning (DRL) methods such as Deep Q-Networks and Hindsight Experience Replay to optimize therapeutic strategies over time. By incorporating multi-modal patient representations, including genomic mutation profiles and clinical phenotypic features, the framework aims to enable personalized treatment planning and dynamic adjustment of therapy protocols. Although currently theoretical, the model provides a foundation for future computational oncology systems that may improve precision medicine through AI-driven decision support and adaptive optimization of cancer treatment pathways.
Branislav Čeperković (Thu,) studied this question.