SMDI: An Index for Measuring Subgingival Microbial Dysbiosis

Area of Science:

• Microbiome research
• Oral microbiology
• Machine learning applications in biology

Background:

• A need exists for a clinically relevant metric to quantify subgingival microbial dysbiosis.
• Existing microbiome analyses lack a simple summary statistic for periodontitis assessment.

Purpose of the Study:

• To develop and validate an intuitive Subgingival Microbial Dysbiosis Index (SMDI).
• To utilize machine learning on 16S rRNA gene sequencing data to identify key microbial species associated with periodontitis and health.

Main Methods:

• Machine learning (random forest) applied to published 16S microbiome data (training and test sets).
• Identification of discriminating species (DS) between periodontitis and healthy states.
• Calculation of SMDI based on the centered log-ratio abundance of DS.
• Diagnostic accuracy assessed using receiver operating characteristic analysis.

Main Results:

• An SMDI based on 49 DS achieved high diagnostic accuracy (AUC 0.96 training, 0.92 test).
• The index effectively differentiates between periodontitis and healthy samples, ranging from -6 (normobiotic) to 5 (dysbiotic).
• Identified key species like *Treponema denticola* (dysbiosis) and *Actinomyces naeslundii* (health); nitrate demonstrated dysbiosis-lowering effects.

Conclusions:

• The developed SMDI is nonbiased, reproducible, and easy to interpret.
• SMDI can identify periodontitis risk, assess treatment response, and serve as a tool in microbiome modulation studies.
• A simplified SMDI using 3 genera shows comparable accuracy.