Assinaturas microbianas orais-fecais nas doenças periodontais: uma análise integrativa de modelagem classificatória supervisionada

Abstract

Alterations in the oral and gut microbiota have been associated with distinct periodontal conditions. However, integrating microbiome data from independent studies remains a significant analytical challenge. This study evaluated the classification of periodontal status using 16S rRNA sequencing data from six distinct cohorts comprising dental biofilm and fecal samples. Following centered log-ratio transformation and batch correction, Random Forest models were trained on combined datasets, with additional batch-stratified analyses used to assess technical variation. In dental biofilm data, overall classification accuracy was moderate (63%) with a multiclass AUC of 0.78, although model performance varied across sequencing batches (AUC 0.94 vs. 0.79). Stool-based models showed similar instability, mainly due to clinical group imbalance and batch origin. Regarding oral taxa, species such as Mogibacterium timidum, Pseudoramibacter alactolyticus, and Aggregatibacter actinomycetemcomitans were associated with more advanced clinical disease, whereas Actinomyces spp. and Veillonella spp. were linked to periodontal health. In fecal samples, gingivitis was associated with genera such as Fusicatenibacter, Blautia, and Dorea, and more severe cases with members of the Lachnospiraceae family. A notable finding was the detection of oral-associated genera (Fusobacterium, Streptococcus, Treponema) in the gut microbiota of gingivitis patients in one sequencing batch, suggesting a possible oral–gut microbial overlap. However, abundance patterns of key taxa such as Prevotella multiformis and Desulfobulbus HMT-041 varied across batches. These findings demonstrate both the potential and limitations of identifying microbiome-based biomarkers across periodontal conditions using integrated datasets, and underscore the importance of methodological harmonization and batch-aware validation in multi-study microbiome research.