Machine learning reveals microbiome differences by periodontitis severity.

Area of Science:

• Oral microbiology
• Computational biology
• Periodontal disease research

Background:

• Periodontitis is a chronic inflammatory disease linked to microbial imbalance.
• Understanding specific microbial signatures for periodontitis severity is crucial but incomplete.

Purpose of the Study:

• To investigate subgingival microbial composition changes across periodontitis severity levels.
• To evaluate the efficacy of machine learning models in classifying periodontitis severity based on microbiome data.

Main Methods:

• 16S rRNA gene sequencing of subgingival plaque from 84 patients.
• Application of five machine learning models (Random Forest, XGBoost) for classification.
• Validation across three external cohorts with feature importance analysis.

Main Results:

• Machine learning models, particularly Random Forest and XGBoost, achieved high classification performance (AUC 0.98).
• Microbial diversity showed a non-significant decreasing trend with increasing disease severity.
• External validation revealed performance variability due to population and methodological differences.
• Fusobacterium, Campylobacter, Stomatobaculum, Leptotrichia, and Segatella were identified as key bacterial contributors.

Conclusions:

• Microbiome-based machine learning models show significant potential for classifying periodontitis severity.
• Robust feature selection and diverse population data are essential for generalizable models.
• Identifying key bacterial taxa advances understanding of periodontal dysbiosis and disease progression.