# Supragingival Biomarker flora of Children With and Without Cariogenic Disease and Black Stains, Aged 3 to 6 Years

## TL;DR

This study examines the oral bacteria in children aged 3 to 6 to identify potential biomarkers for dental caries and black stains.

## Contribution

The study identifies specific bacteria and machine learning models that could serve as biomarkers for caries and black stains in young children.

## Key findings

- Microbial diversity was similar across groups, but specific bacteria varied in abundance.
- Seven machine learning models showed high accuracy in predicting caries and black stains.
- Keystone bacteria in caries and black stain groups could act as potential biomarkers.

## Abstract

The oral microbiome plays a pivotal role in the occurrence and progression of dental caries and black stain (BS) pigment.

The aim of this study was to explore the keystone microbiota and potential biomarkers of caries and BS pigment in 3 to 6-year-old children.

A total of 122 children were included, namely, healthy controls (HC, n = 32), those with severe early childhood caries (SECC, n = 31), those with BS pigment but caries-free (BSCF, n = 29), and those with SECC and BS pigment (SECCBS, n = 30). Supragingival plaques were collected for 16S rRNA sequencing followed by bioinformatics analysis.

Seven phyla and 14 genera were identified in all the samples, and differences in relative abundance were observed. Alpha diversity analysis revealed that the richness and diversity of the bacterial communities were similar across the HC, BSCF, SECC and SECCBS groups (P > .05). Different bacterial species were identified in the six paired groups (P < .05). With respect to the disparities in keystone nodes, the SECC group had the highest value of 66, followed by the SECCBS and BSCF groups and the HC group (56, 47 and 33, respectively). The areas under roc curve for the 10 machine learning models were systematically evaluated, and seven models yielded exceptional results, including support vector machine (SVM)-linear and SVM-RBF for BSCF–SECC, naïve Bayes classification for BSCF–SECCBS, decision trees for HC–BSCF, LASSO for HC–SECC, and SVM-poly for HC–SECCBS and K nearest neighbour for SECC–SECCBS.

The diversity of the microbial community has little influence on the development of dental caries and black staining. However, specific bacteria exhibited different relative abundances across the HC, SECC, BSCF, and SECCBS groups; therefore, those bacteria may serve as candidate biomarkers. Co-occurrence network approaches and differential machine learning models can be used to predict a spectrum of dental caries in primary dentition, providing a convenient and preventive strategy.