面向药品重定位的医学知识图谱的构建与应用

面对突然爆发的未知病毒，从头开始研发新药花费太长时间，“老药新用”的药物重定位策略是最佳选择。综合全面的医学知识库是利用计算机进行药品重定位的基础，但目前的药品重定位研究既缺乏综合各类药品知识的大规模知识库，也没有利用大规模知识库的知识表达手段。针对上述两个问题，本项目将知识图谱引入药品重定位，研究基于知识图谱的药品重定向方法。本项目在整合多个国外主流医药知识库的基础上，融合中文药品知识库，构建综合各类药品知识的医药知识图谱。同时从最新的医学文献和专利抽取知识，不断扩充知识图谱。在建立知识图谱后，针对医药知识图谱的特点，改进目前的表示学习模型，得到知识图谱的嵌入表示，采用PU学习和链接预测的方法进行药品重定位。本项目可为应急条件下的药品快速研发提供支持，提高我国应对紧急公共卫生事件的能力；还能辅助药品研发，有效降低研发经费和失败风险，提高药物研发的投入产出比。

When dealing with a sudden outbreak of an unknown virus, it takes too long to develop new drugs from scratch, and the drug repositioning strategy “using old drugs for new diseases” is the best choice. The comprehensive medical knowledge base is the basis of in silico drug repositioning. However, the current research on drug repositioning is lack of large-scale knowledge base integrating all kinds of drug knowledge and knowledge representation methods for large-scale knowledge base. To address the above two problems, we introduce knowledge graph which has large-scale network storage and representation capabilities into drug repositioning. We construct a medical knowledge graph, and use representation learning to perform drug repositioning. We integrate multiple foreign medical knowledge bases, Chinese drug knowledge bases, to build a medical knowledge map that includes various types of drug knowledge. At the same time, we extract knowledge from the latest medical literature and patents, and constantly expand the knowledge graph. According to the characteristics of the medical knowledge graph, we improve the current representation learning model to obtain the embedded representation of the knowledge graph and employ PU learning and link prediction to implement drug repositioning. The drug knowledge graph and drug repositioning method researched by this project can provide support for rapid drug development under emergency conditions, improve China's ability to respond to emergency public health incidents. Our research can also assist in drug research, effectively reduce research and development funding and failure risks, and increase the input-output ratio of drug research and development under the condition of relatively limited scientific research resources.