.png){kind=icon}
.png){kind=icon}
Return
Oral microbiome classification of elevated salivary glucose in a large adolescent Kuwaiti population
.png){kind=icon}
OA
AI
C
A
A
F
H
H
H
DOI：10.3389/fmicb.2026.1863859.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
BackgroundThe oral cavity harbors a complex microbial ecosystem that reflects both local and systemic health. Metabolic disorders; including impaired glucose regulation; are known to alter the oral microbiome; however; whether oral microbial signatures can classify elevated salivary glucose status in large; healthy pediatric populations has received limited investigation.ObjectiveTo determine whether oral microbiome composition; assessed by DNA probe analysis; can classify elevated salivary glucose status in Kuwaiti adolescents; and to identify the bacterial taxa contributing most to this classification.MethodsOral microbiome data from 8; 173 adolescents (mean age 10.0 ± 0.7 years) were analyzed using relative abundances of 42 bacterial species measured by the Socransky checkerboard DNA–DNA hybridization method. Salivary glucose was dichotomized at the cohort median (0.148 mg/dL) to define Low Glucose (≤ median; n = 4; 087) and High Glucose (> median; n = 4; 086) groups. Alpha diversity (Shannon and Simpson indices) was compared between groups using Mann–Whitney U tests; and beta diversity (Bray–Curtis dissimilarity) was assessed using PERMANOVA. Following removal of highly correlated features (Pearson r > 0.80); a logistic regression model with L1 regularization and 5-fold stratified cross-validation was trained and evaluated.ResultsBoth alpha diversity indices were notably lower in the High Glucose group (p < 0.001). PERMANOVA confirmed a small but statistically significant compositional shift between groups (pseudo-F = 35.43; R2 = 0.0174; p = 0.001). The classification model achieved an accuracy of 75.3%; sensitivity of 73.9%; specificity of 76.7%; and ROC AUC of 0.820. All 37 bacterial features retained after correlation filtering received non-zero L1 coefficients. Fusobacterium nucleatum subsp. vincentii (coefficient = +3.624) was the taxon most strongly positively associated with the High Glucose group; while Aggregatibacter actinomycetemcomitans (coefficient = −0.882) was the taxon most strongly inversely (negatively) associated with the High Glucose group.ConclusionOral microbiome composition can discriminate between adolescents with elevated and normal salivary glucose levels with moderate discriminatory accuracy under internal cross-validation. Reduced microbial diversity and specific bacterial signatures characterize the high-glucose state; supporting the potential utility of oral microbiome profiling as a non-invasive strategy for early metabolic risk identification in pediatric populations.
Keywords:
machine learning
adolescents
microbial diversity
oral microbiome
Fusobacterium nucleatum
logistic regression
metabolic risk
salivary glucose
Journal
IF:
4.5
Papers:
4.0W
Citations:
16.6W
Organization
Cited Papers