浙江中医药大学学报2025，Vol.49Issue(12)：1536-1551,16.DOI:10.16466/j.issn1005-5509.2025.12.005
基于知识图谱的经方类方衍化规律挖掘
Mining the Evolution Rules of Classical Prescriptions Similar in Medication Based on the Completion of Knowledge Graph——A Case Study of Six Clustered Formula Categories
摘要
Abstract
[Objective]To enhance the missing and ambiguous contents of classical prescriptions in Treatise on Cold Pathogenic and Miscellaneous Diseases through knowledge graph completion techniques,and explore the evolution patterns of classical prescriptions similar in order to provide theoretical support for clinical application and the study of classical prescriptions.[Methods]Based on the established knowledge graph of classical prescriptions,this research adopts two knowledge completion techniques,namely symbol-based representation and vector-based representation,to complete the contents of classical prescriptions similar in Treatise on Pathogenic and Miscellaneous Disease.The specific methods include initially screening drug completions using both symbol-based and vector-based representation methods,then fusing the completion results of the two methods to obtain completed triplets,which are then traversed in the classical prescription database to derive completed prescriptions.It compares and screens the supplemented prescriptions against the original prescriptions,and introduces independent review and evaluation by experts in the field of traditional Chinese medicine.Finally,clustering analysis techniques are used to explore the evolution patterns of classical prescriptions.[Results]This study systematically expanded the classical prescription knowledge graph of Treatise on Cold Pathogenic and Miscellaneous Diseases by integrating symbolic and vector representation techniques.The symbolic representation completion enriched the knowledge graph with 19 relationship types(totaling 10 085 triplets),including 4 233"containment"and 3 407"therapeutic"relationships,Which validated through Neo4j graph database.Vector representation completion using TransE model further optimized the graph,increasing total relationships to 10 468(with 4 616"containment"relationships)and adding 383 new triplets.Fusion screening of 250 classical prescriptions from Song edition Treatise on Cold Pathogenic Diseases and Synopsis of the Golden Chamber yielded 333 additional formulas(total 644 formulas)and 239 medicinal entities after expert validation.The completed knowledge graph comprised 24 entity types(2 455 nodes),with cluster analysis revealing 6 evolutionary patterns of classical formulas.[Conclusion]This study utilizes knowledge graph completion techniques,combining both symbol-based and vector-based completion methods,to enhance the knowledge graph of classical prescriptions,revealing previously implicit knowledge of classical prescriptions and potential new prescriptions,and exploring the evolution patterns of classical prescriptions through clustering analysis.关键词
经方/类方/人工智能/知识图谱/知识补全/知识挖掘/图论/《伤寒杂病论》Key words
classical prescriptions/similar prescriptions/artificial intelligence/knowledge graph/knowledge completion/knowledge mining/graph theory/Treatise on Cold Pathogenic and Miscellaneous Disease分类
医药卫生引用本文复制引用
喻滔,薛江涛,陈可玥,李湛,曹灵勇,林树元..基于知识图谱的经方类方衍化规律挖掘[J].浙江中医药大学学报,2025,49(12):1536-1551,16.基金项目
国家中医药管理局科技司-浙江中医药管理局共建科技计划项目(GZY-ZJ-KJ-23018) （GZY-ZJ-KJ-23018）
浙江中医药大学横向(涉企)项目(2022-HT-837) （涉企）
浙江省"尖兵领雁+x"科技计划项目(2025C02194) Science and Technology Department of the State Administration of Traditional Chinese Medicine-Zhejiang Administration of Traditional Chinese Medicine Joint Construction Science and Technology Project(GZY-ZJ-KJ-23018) （2025C02194）
Zhejiang Chinese Medical University Horizontal(Enterprise-Related)Project(2022-HT-837) （Enterprise-Related）
Zhejiang Provincial"Pioneer&Leading Goose"R&D Program(2025C02194) （2025C02194）
