EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Using gut microbiome metagenomic hypervariable features for diabetes screening and typing through supervised machine learning

Xavier Chavarria, Hyun Seo Park, Singeun Oh, Dongjun Kang, Jun Ho Choi, Myungjun Kim, Yoon Hee Cho, Myung-Hee Yi, Ju Yeong Kim

Microbial Genomics · 2025

DOI: 10.1099/mgen.0.001365 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Microbial Genomics, 2025. Verfasst von Xavier Chavarria und 8 weiteren Autor:innen. Beteiligte Einrichtung: Department of Tropical Medicine, Institute of Tropical Medicine, Arthropods of Medical Importance Resource Bank, Yonsei University College of Medicin…. Hauptthemen laut MEDLINE: Gastrointestinal Microbiome, Supervised Machine Learning, Diabetes Mellitus, Type 2, Metagenomics. Die Arbeit nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Foundation, Korea Health Industry Development Institute, National Institute of Biological Resources.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Diabetes mellitus is a complex metabolic disorder and one of the fastest-growing global public health concerns. The gut microbiota is implicated in the pathophysiology of various diseases, including diabetes. This study utilized 16S rRNA metagenomic data from a volunteer citizen science initiative to investigate microbial markers associated with diabetes status (positive or negative) and type (type 1 or type 2 diabetes mellitus) using supervised machine learning (ML) models. The diversity of the microbiome varied according to diabetes status and type. Differential microbial signatures between diabetes types and negative group revealed an increased presence of Brucellaceae , Ruminococcaceae , Clostridiaceae , Micrococcaceae , Barnesiellaceae and Fusobacteriaceae in subjects with diabetes type 1, and Veillonellaceae , Streptococcaceae and the order Gammaproteobacteria in subjects with diabetes type 2. The decision tree, elastic net, random forest (RF) and support vector machine with radial kernel ML algorithms were trained to screen and type diabetes based on microbial profiles of 76 subjects with type 1 diabetes, 366 subjects with type 2 diabetes and 250 subjects without diabetes. Using the 1000 most variable features, tree-based models were the highest-performing algorithms. The RF screening models achieved the best performance, with an average area under the receiver operating characteristic curve (AUC) of 0.76, although all models lacked sensitivity. Reducing the dataset to 500 features produced an AUC of 0.77 with sensitivity increasing by 74% from 0.46 to 0.80. Model performance improved for the classification of negative-status and type 2 diabetes. Diabetes type models performed best with 500 features, but the metric performed poorly across all model iterations. ML has the potential to facilitate early diagnosis of diabetes based on microbial profiles of the gut microbiome.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Gastrointestinal MicrobiomeSupervised Machine LearningDiabetes Mellitus, Type 2MetagenomicsDiabetes Mellitus, Type 1BacteriaHumansRNA, Ribosomal, 16SMaleFemaleMiddle AgedAdultMachine LearningMetagenomeSupport Vector Machine

Stichwörter der Autor:innen

diabetes mellitusgut microbiomemetabarcodingmicrobial markersrandom forestsupervised machine learning

Substanzen

RNA, Ribosomal, 16S

Autor:innen und Institutionen

  • Department of Tropical Medicine, Institute of Tropical Medicine, Arthropods of Medical Importance Resource Bank, Yonsei University College of Medicine, Yonsei-ro 50-1, Seodaemun-gu, Seoul 03722, Republic of Korea.
  • Department of Systems Biology, Yonsei University College of Life Science and Biotechnology, Yonsei-ro 50-1, Seodaemun-gu, Seoul 03722, Republic of Korea.

Förderung

  • National Research Foundation
  • Korea Health Industry Development Institute
  • National Institute of Biological Resources

Zitieren

APA

Chavarria, X., Park, H. S., Oh, S., Kang, D., Choi, J. H., Kim, M., Cho, Y. H., Yi, M. H., & Kim, J. Y. (2025). Using gut microbiome metagenomic hypervariable features for diabetes screening and typing through supervised machine learning. Microbial Genomics. https://doi.org/10.1099/mgen.0.001365

Vancouver

Chavarria X, Park HS, Oh S, Kang D, Choi JH, Kim M, et al. Using gut microbiome metagenomic hypervariable features for diabetes screening and typing through supervised machine learning. Microbial Genomics. 2025. doi:10.1099/mgen.0.001365

BibTeX
@article{Chavarria2025,
  author = {Chavarria, Xavier and Park, Hyun Seo and Oh, Singeun and Kang, Dongjun and Choi, Jun Ho and Kim, Myungjun and Cho, Yoon Hee and Yi, Myung-Hee and Kim, Ju Yeong},
  title = {Using gut microbiome metagenomic hypervariable features for diabetes screening and typing through supervised machine learning},
  journal = {Microbial Genomics},
  year = {2025},
  doi = {10.1099/mgen.0.001365}
}
Kostenlos starten →