A Hybrid GNN-RL Framework for Drug Design: A Case Study on Type 2 Diabetes

جاري التحميل...
صورة مصغرة

التاريخ

عنوان الدورية

ردمد الدورية

عنوان المجلد

الناشر

University of Science and Technology, Sana'a

خلاصة

The sheer scale of the chemical space makes modern drug discovery an incredibly resource-intensive endeavor to optimize multiple pharmacological and safety constraints simultaneously. Even though the AI and ML have enabled computational support for specific tasks, most existing approaches focused on the isolated problems such as molecular property prediction or molecule generation. This study presents a modular end-to-end AI framework that integrates Graph Neural Networks (GNNs) with Reinforcement Learning (RL) for de novo drug design. In our methodology the Molecules are represented as graphs and encoded into embeddings using GNNs, which are then used as states for an RL agent to guide molecular optimization. The framework is applied to Type 2 Diabetes Mellitus (T2DM), targeting well-known proteins such as DPP-4, SGLT2, GLP-1, and PI3K, using publicly available datasets from ChEMBL and DrugBank. Evaluation Shifting the focus toward component interaction, system-level behavior, and full pipeline execution. The results confirm the feasibility of coupling GNN-based molecular representations with RL optimization in a unified and scalable framework, providing a strong foundation for future AI-driven drug discovery workflows.

الوصف

اقتباس

Khorees, A. K. Y. (2026). A Hybrid GNN-RL Framework for Drug Design: A Case Study on Type 2 Diabetes [Master's thesis, University of Science and Technology, Sana'a].

Endorsement

Review

Supplemented By

Referenced By