Drug discovery remains a slow and costly process, limiting the rapid development of effective cancer therapies. This study presents a computational framework that applies Deep Reinforcement Learning (DRL) to generate novel molecules targeting the Epidermal Growth Factor Receptor (EGFR), a key cancer related protein. Bioactive compounds and molecular data were retrieved from ChEMBL and represented in Simplified Molecular Input Line Entry System (SMILES) format. Molecular descriptors were extracted using RDkit, and a DRL model (Proximal Policy Optimization) was trained to propose drug candidates optimized for EGFR binding. Generated molecules were evaluated through molecular docking using AutoDock Vina and Absorption, Distribution, Metabolism, Excretion, Toxicity (ADMET) profiles were predicted to assess therapeutic suitability. The top candidate exhibited strong binding affinity (-8.9 kcal/mol), ideal Root Mean Square Deviation (RMSD) (0.0), and favorable druglike properties. Incorporating patient specific data, including mutation type, HLA profile, and disease stage further improved binding affinity, demonstrating the value of personalized molecule optimization. This work demonstrates the potential of AI guided approaches to accelerate early-stage cancer drug discovery and provides a foundation for integrating computational and experimental methods within precision oncology.

DRL EGFR Personalized Medicine Molecular Docking ADMET AI Drug Design Precision Oncology