Supervised machine learning-based classification of oral malodor based on the microbiota in saliva samples

Area of Science:

• Microbiology
• Bioinformatics
• Machine Learning

Background:

• Oral malodor is a common condition often linked to volatile sulfur compounds produced by oral microbiota.
• Accurate detection of oral malodor-causing compounds like methyl mercaptan is crucial for diagnosis and treatment.
• Current methods may require identification of specific bacterial species, which can be complex and time-consuming.

Purpose of the Study:

• To develop and evaluate machine learning models for classifying oral malodor based on salivary microbiota.
• To utilize methyl mercaptan concentrations and terminal restriction fragment length polymorphism (T-RFLP) data for classification.
• To assess the effectiveness of Support Vector Machine (SVM), Artificial Neural Network (ANN), and Decision Tree algorithms.

Main Methods:

• Saliva samples from 309 subjects were analyzed using 16S rRNA gene T-RFLP to obtain fragment length and peak area data.
• Two types of frequencies were calculated: peak area frequency and species-containing sample frequency.
• Supervised machine learning models (SVM, ANN, Decision Tree) were trained on 308 samples to classify methyl mercaptan presence in the remaining subject.

Main Results:

• The SVM model achieved the highest classification accuracy with a specificity of 95.0% and sensitivity of 51.1%.
• The ANN model, when weighted with entropy from frequency of appearance, improved accuracy to 81.9% (sensitivity 60.2%, specificity 90.5%).
• The Decision Tree model demonstrated low classification accuracy across all tested conditions.

Conclusions:

• Machine learning models, particularly SVM, can effectively classify the presence of methyl mercaptan in saliva, indicating oral malodor.
• These models offer a high specificity screening method for oral malodor, potentially reducing the need for specialist clinic visits.
• Classification does not necessitate the identification of specific oral microbiota species responsible for malodor, simplifying the diagnostic process.