Kumari Madhavi, Chauhan Rohit, Garg Prabha
Department of Pharmacoinformatics, National Institute of Pharmaceutical Education and Research (NIPER), S.A.S. Nagar, Sector 67, S.A.S. Nagar, Mohali, Punjab, 160062, India.
Department of Computer Science, National Institute of Technology (NIT), Durgapur, MG Road, Durgapur, West Bengal, 713209, India.
Mol Divers. 2025 Mar 14. doi: 10.1007/s11030-025-11164-z.
Biomedical knowledge graphs have emerged as powerful tools for drug discovery, but existing platforms often suffer from outdated information, limited accessibility, and insufficient integration of complex data. This study presents MedKG, a comprehensive and continuously updated knowledge graph designed to address these challenges in precision medicine and drug discovery. MedKG integrates data from 35 authoritative sources, encompassing 34 node types and 79 relationships. A Continuous Integration/Continuous Update pipeline ensures MedKG remains current, addressing a critical limitation of static knowledge bases. The integration of molecular embeddings enhances semantic analysis capabilities, bridging the gap between chemical structures and biological entities. To demonstrate MedKG's utility, a novel hybrid Relational Graph Convolutional Network for disease-drug link prediction, MedLINK was developed and used in case studies on clinical trial data for disease drug link prediction. Furthermore, a web-based application with user-friendly APIs and visualization tools was built, making MedKG accessible to both technical and non-technical users, which is freely available at http://pitools.niper.ac.in/medkg/.
生物医学知识图谱已成为药物发现的强大工具，但现有平台往往存在信息过时、可访问性有限以及复杂数据整合不足等问题。本研究提出了MedKG，这是一个全面且不断更新的知识图谱，旨在解决精准医学和药物发现中的这些挑战。MedKG整合了来自35个权威来源的数据，涵盖34种节点类型和79种关系。一个持续集成/持续更新管道确保MedKG始终保持最新状态，解决了静态知识库的一个关键限制。分子嵌入的整合增强了语义分析能力，弥合了化学结构与生物实体之间的差距。为了证明MedKG的实用性，开发了一种用于疾病-药物关联预测的新型混合关系图卷积网络MedLINK，并将其用于疾病药物关联预测的临床试验数据案例研究。此外，还构建了一个具有用户友好型应用程序编程接口和可视化工具的基于网络的应用程序，使技术用户和非技术用户都能访问MedKG，该应用程序可在http://pitools.niper.ac.in/medkg/免费获取。
