Skip to content
DraconiticResearch
Search Sources API
223,421 studies31 sourcesUpdated just now
Peer-reviewedCC BY-NC-NDnpj Digital Medicine

MedKGent: a large language model agent framework for constructing temporally evolving medical knowledge graph

  • Duzhen Zhang1
  • Zixiao Wang1
  • Zhongzhi Li2
  • Yahan Yu3
  • Shuncheng Jia2
  • Jiahua Dong1
  • Haotian Xu4
  • Xing Wu2
  • Yingying Zhang5
  • Tielin Zhang6,2
  • Jie Yang7
  • Xiuying Chen1
1 more authors
  • Le Song8,1
8 affiliations
  1. 1Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE
  2. 2University of Chinese Academy of Sciences, Beijing, China
  3. 3Kyoto University, Kyoto, Japan
  4. 4Tsinghua University, Beijing, China
  5. 5East China Normal University, Shanghai, China
  6. 6Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai, China
  7. 7Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA
  8. 8GenBio AI, San Francisco, USA
Open at sourcePDFDOICite

Abstract

The rapid expansion of medical literature challenges the scalable structuring of domain knowledge. Knowledge Graphs (KGs) offer a solution, yet current construction methods lack generalizability and ignore the temporal dynamics of evolving knowledge. To address this, we introduce MedKGent, a Large Language Model (LLM) agent framework for building temporally evolving medical KGs. Using over 10 million PubMed abstracts from 1975 to 2023, MedKGent incrementally constructs a KG daily via two specialized agents. The Extractor Agent identifies knowledge triples and assigns confidence scores, while the Constructor Agent integrates these triples into a temporal graph, reinforcing recurring knowledge and resolving conflicts. The resulting KG contains 156,275 entities and 2,971,384 triples, making it, to our knowledge, the largest LLM-derived medical KG to date. Automated and expert assessments showed triple-validity rates approaching 90%. In downstream evaluations, MedKGent-KG significantly improved retrieval-augmented generation for five LLMs across seven medical question-answering benchmarks. Together, these results position MedKGent as a scalable and temporally aware infrastructure for medical knowledge representation and literature-grounded AI research.

Related

  1. Knowledge distillation-based lightweight generative large language models for standardized quality control of Chinese radiology reports
    Duoning Jiang, Xiao Han, Ke Xu +1
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textEpidemiology & public healthCrossref
  2. A Knowledge Database of Physical Activity Interventions for Hypertension using Large Language Models
    C. Li, C. Obioha, X. Cui +2
    medRxivPreprintCC BYFull textEpidemiology & public healthmedRxiv
  3. Development of knowledge graphs for emergency events utilizing pre-trained language models
    Heng Yang, Guangzu Bai, Nan Wang +1
    Scientific ReportsPeer-reviewedCC BYFull textGeneral medicineDOAJ
  4. Recent Advances in Knowledge Representation and Management in Healthcare: the Emerging Role of Large Language Models
    Catalina Martínez Costa, Jean Charlet, Licong Cui
    Yearbook of Medical InformaticsPeer-reviewedCC BY 4.0Full textEpidemiology & public healthPMC
  5. Knowledge-Guided Large Language Model for Automatic Pediatric Dental Record Understanding and Safe Antibiotic Recommendation
    Zihan Han, Zhitao Shu, Caifeng Li
    Fundamental Scientific Reports in Multidisciplinary AreasPeer-reviewedCC BYFull text19 citedEpidemiology & public healthOpenAlex

Cite

BibTeXRIS
BibTeX
@article{zhang2026medkgent,
  title = {{MedKGent: a large language model agent framework for constructing temporally evolving medical knowledge graph}},
  author = {Duzhen Zhang and Zixiao Wang and Zhongzhi Li and Yahan Yu and Shuncheng Jia and Jiahua Dong and Haotian Xu and Xing Wu and Yingying Zhang and Tielin Zhang and Jie Yang and Xiuying Chen and Le Song},
  journal = {npj Digital Medicine},
  year = {2026},
  month = aug,
  doi = {10.1038/s41746-026-03058-7},
  url = {https://doi.org/10.1038/s41746-026-03058-7},
  note = {CC BY-NC-ND},
}
RIS
TY  - JOUR
TI  - MedKGent: a large language model agent framework for constructing temporally evolving medical knowledge graph
AU  - Duzhen Zhang
AU  - Zixiao Wang
AU  - Zhongzhi Li
AU  - Yahan Yu
AU  - Shuncheng Jia
AU  - Jiahua Dong
AU  - Haotian Xu
AU  - Xing Wu
AU  - Yingying Zhang
AU  - Tielin Zhang
AU  - Jie Yang
AU  - Xiuying Chen
AU  - Le Song
PY  - 2026
DA  - 2026/08/04
JO  - npj Digital Medicine
DO  - 10.1038/s41746-026-03058-7
UR  - https://doi.org/10.1038/s41746-026-03058-7
AB  - The rapid expansion of medical literature challenges the scalable structuring of domain knowledge. Knowledge Graphs (KGs) offer a solution, yet current construction methods lack generalizability and ignore the temporal dynamics of evolving knowledge. To address this, we introduce MedKGent, a Large Language Model (LLM) agent framework for building temporally evolving medical KGs. Using over 10 million PubMed abstracts from 1975 to 2023, MedKGent incrementally constructs a KG daily via two specialized agents. The Extractor Agent identifies knowledge triples and assigns confidence scores, while the Constructor Agent integrates these triples into a temporal graph, reinforcing recurring knowledge and resolving conflicts. The resulting KG contains 156,275 entities and 2,971,384 triples, making it, to our knowledge, the largest LLM-derived medical KG to date. Automated and expert assessments showed triple-validity rates approaching 90%. In downstream evaluations, MedKGent-KG significantly improved retrieval-augmented generation for five LLMs across seven medical question-answering benchmarks. Together, these results position MedKGent as a scalable and temporally aware infrastructure for medical knowledge representation and literature-grounded AI research.
SN  - 2398-6352
ER  -