Beyond the Red Complex: De Novo Marker Discovery Uncovers Novel Periodontitis-Associated Taxa and Enables Non-Invasive Machine Learning Diagnosis

Periodontitis affects over 1 billion people worldwide, but its diagnosis relies on clinical measures that do not capture underlying microbial dysbiosis. A research team analyzed 341 microbiome samples from supragingival and subgingival sites (218 with periodontitis, 123 healthy) from nine countries using a reference-free MetaMarker method. They identified 2,142 significant markers, including 1,999 associated with periodontitis and 143 with health, with 128 markers (6.4%) being novel, including an uncultivated Paludibacteraceae genus. Machine learning models, particularly XGBoost and gradient boosting, achieved AUC=0.96 on external validation cohorts. A compact panel of 20 markers from five taxa successfully reproduced the results of the full marker set. This non-invasive diagnostic method has potential for clinical application.