Related Experiment Video
Updated: May 2, 2026
10:42
Oral Biofilm Sampling for Microbiome Analysis in Healthy Children
Published on: December 31, 2017
18.0K
Prediction model for periodontitis stage based on the salivary microbiome
Jaewoong Lee1, Hyun-Joo Kim2,3, Eun-Hye Kim4,5
1Department of Biomedical Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan, Republic of Korea.
Msystems
|March 11, 2026
Summary
Salivary microbiome analysis can classify periodontitis stages using machine learning. This non-invasive method identifies key bacteria like Actinomyces spp. for early and accurate diagnosis of periodontal disease.
Area of Science:
- Oral microbiology
- Computational biology
- Periodontology
Background:
- Periodontitis is a prevalent oral disease with significant health implications.
- Accurate and early diagnosis is crucial for effective treatment but challenging with traditional methods.
- Salivary microbiome alterations are linked to periodontal health status.
Purpose of the Study:
- To characterize salivary microbiome compositions for classifying periodontal health and periodontitis stages.
- To develop a machine learning model for non-invasive periodontitis diagnosis.
- To identify key microbial taxa associated with different stages of periodontitis.
Main Methods:
- 16S ribosomal RNA gene sequencing of saliva samples from 250 subjects (100 healthy, 150 periodontitis).
- Analysis of alpha diversity and identification of differentially abundant taxa using ANCOM.
- Development of Random Forest machine learning models for classification of periodontitis stages.
Main Results:
- Significant differences in salivary microbiome alpha diversity between healthy and periodontitis groups.
- Identification of 20 differentially abundant taxa, with Porphyromonas gingivalis and Actinomyces spp. being key classifiers.
- Random Forest model achieved high accuracy in classifying periodontitis stages (AUC 0.829), early-stage disease (AUC 0.736), and distinguishing from healthy individuals (AUC 0.924).
- Model demonstrated good performance on external datasets, with minor variations potentially due to ethnicity.
Conclusions:
- Salivary microbiome profiles can accurately classify periodontal health and various stages of periodontitis.
- A machine learning model utilizing key microbial taxa offers a non-invasive diagnostic tool for periodontitis.
- This approach supports precise, accessible, and early diagnosis, potentially improving dental disease management.
