Clinical pathogen adjudication in metagenomic sequencing using a large language model

Abstract

Clinical metagenomic sequencing detects hundreds of organisms whose relevance to infection requires clinicians manual interpretation. We retrospectively reviewed 4,855 specimens from 19 sampling sites, linked 1.33 million detected organisms to clinical culprit-pathogen determinations, and fine-tuned a large language model on these labels. The model recovered 92.44% of reference culprit calls in the held-out test set and 84.49% in a later cohort. In a blinded re-adjudication of 215 cases, the model matched 90.5% of culprit calls on which both expert reviewers agreed. Adjudication accuracy varied with organism identity and specimen types. mNGS-derived metrics such as read counts or genome coverage alone were insufficient to identify culprit pathogens, and organisms classified as common colonizers could still be judged causative. Certain organisms were frequently co-reported as causative in the same specimen. These findings support further evaluation of human and AI-assisted report review, particularly regarding missed pathogens, excess calls, and differences across specimen types.

Metagenomic next-generation sequencing in tuberculous periprosthetic joint infection: pathogen-detection yield and clinical characteristics compared with common bacterial PJI.
Jiayu Li, Yuntao Liao, Xiaolin Li +7
Frontiers in cellular and infection microbiology7 Oct 2026Peer-reviewedCC BYFull textSurgery & anaesthesiaEurope PMC
Wastewater metagenomic sequencing enables broad pathogen and resistome monitoring in Lagos, Nigeria
E. Parker, F. Oladipo, J. I. Levy +5
medRxiv22 Sep 2026PreprintCC BY-NDFull textEpidemiology & public healthmedRxiv
Comparative evaluation of metagenomic and targeted next-generation sequencing for pathogen detection in bronchoalveolar lavage fluid from patients with lower respiratory tract infections: a retrospective cohort study.
Rui Zhao, Heya Na, Wei Miao +5
Frontiers in cellular and infection microbiology14 Oct 2026Peer-reviewedCC BYFull textGeneral medicineEurope PMCPMC
Metagenomic Next-Generation Sequencing Enhances Pathogen Detection and Prognostic Stratification in Hematopoietic Stem Cell Transplantation Recipients.
Ying Liu, Xintong Bian, Hongxu Wang +1
Infection and drug resistance23 Oct 2026Peer-reviewedCC BY-NCFull textGeneral medicineEurope PMCPMC
Metagenomic next-generation sequencing of matched donor–recipient pairs reveals pathogen transmission dynamics and informs antimicrobial therapy in liver transplantation: a retrospective cohort study
Shuai Feng, Jinquan Liu, Xiaojing Zhang +7
Frontiers in cellular and infection microbiology19 Oct 2026Peer-reviewedCC BYFull textGeneral medicineEurope PMCPMCEurope PMC

Cite

BibTeX RIS

BibTeX

@misc{wei2026clinical,
  title = {{Clinical pathogen adjudication in metagenomic sequencing using a large language model}},
  author = {M. Wei and Y. Zhao and C. Fang and S. Zhang and K. Xu and S. Li and L. Liu and Y. Pang and L. Liang and Z. Deng and M. Li},
  howpublished = {medRxiv},
  year = {2026},
  month = oct,
  doi = {10.64898/2026.10.16.26361411},
  url = {https://www.medrxiv.org/content/10.64898/2026.10.16.26361411v1},
  note = {CC BY-NC-ND},
}

RIS

TY  - UNPB
TI  - Clinical pathogen adjudication in metagenomic sequencing using a large language model
AU  - M. Wei
AU  - Y. Zhao
AU  - C. Fang
AU  - S. Zhang
AU  - K. Xu
AU  - S. Li
AU  - L. Liu
AU  - Y. Pang
AU  - L. Liang
AU  - Z. Deng
AU  - M. Li
PY  - 2026
DA  - 2026/10/16
PB  - medRxiv
DO  - 10.64898/2026.10.16.26361411
UR  - https://www.medrxiv.org/content/10.64898/2026.10.16.26361411v1
AB  - Clinical metagenomic sequencing detects hundreds of organisms whose relevance to infection requires clinicians manual interpretation. We retrospectively reviewed 4,855 specimens from 19 sampling sites, linked 1.33 million detected organisms to clinical culprit-pathogen determinations, and fine-tuned a large language model on these labels. The model recovered 92.44% of reference culprit calls in the held-out test set and 84.49% in a later cohort. In a blinded re-adjudication of 215 cases, the model matched 90.5% of culprit calls on which both expert reviewers agreed. Adjudication accuracy varied with organism identity and specimen types. mNGS-derived metrics such as read counts or genome coverage alone were insufficient to identify culprit pathogens, and organisms classified as common colonizers could still be judged causative. Certain organisms were frequently co-reported as causative in the same specimen. These findings support further evaluation of human and AI-assisted report review, particularly regarding missed pathogens, excess calls, and differences across specimen types.
ER  -

Abstract

Related

Cite