Comparative machine learning analysis of saliva and plaque microbiomes in Kuwaitis with type 1 diabetes

Area of Science:

• Microbiome research
• Pediatric endocrinology
• Immunology

Background:

• Type 1 diabetes (T1D) is linked to microbial imbalances.
• Research on the oral microbiome's role in T1D is limited.
• Oral and gut microbiomes interact, influencing systemic health.

Purpose of the Study:

• To compare salivary and dental plaque microbiomes for classifying pediatric T1D status.
• To determine which oral niche is more strongly associated with T1D.
• To assess the predictive performance of oral microbiome data in T1D diagnosis.

Main Methods:

• Collected paired saliva and plaque samples from 46 children (23 T1D, 23 controls).
• Sequenced 16S rRNA gene for microbial profiling.
• Utilized QIIME 2 for data processing and random forest models for classification.

Main Results:

• Saliva showed higher bacterial load and abundance than plaque.
• Saliva-based models achieved 94.2% accuracy in classifying T1D, outperforming plaque models (73.3%).
• Saliva models demonstrated superior discriminative performance (AUC ~0.94 vs. 0.75 for plaque).

Conclusions:

• The salivary microbiome is a more effective biospecimen than dental plaque for T1D characterization in children.
• Saliva offers superior classification accuracy and sensitivity for T1D status.
• Saliva holds potential as a non-invasive tool for microbiome-based T1D monitoring.