Machine Learning Models For Predicting Cad By Porphyromonas Gingivalis Concentration

To investigate the independent predictive value of Porphyromonas gingivalis (Pg) concentration for coronary artery disease (CAD), construct a Pg concentration-integrated risk prediction model, and evaluate its potential as an early-warning biomarker. Three machine learning (ML) algorithms—Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM)—were employed to develop CAD risk prediction models. Model performance was validated using the area under the receiver operating characteristic curve (AUC), Brier score, and decision curve analysis (DCA). The independent predictive contribution of Pg concentration was quantified via net reclassification improvement (NRI) and integrated discrimination improvement (IDI). A nomogram was developed for visualization of risk assessment. The RF model demonstrated optimal performance (AUC=0.965, Brier score=0.102), significantly outperforming the XGBoost model (AUC=0.859, Brier score=0.169) and SVM model (AUC=0.148, Brier score=0.302). SHAP analysis identified Pg concentration as the key predictive factor. Its inclusion significantly enhanced model classification efficacy (NRI=0.444, 95% CI=0.134–0.755, P<0.05; IDI=0.444, 95% CI=0.273–0.615, P<0.05). DCA curves confirmed the model’s clinical applicability across a broad range of risk thresholds. This study successfully established an ML-based CAD prediction model using Pg concentration, validating its independent predictive influence on CAD risk stratification. The model provides a novel microbial biomarker strategy for early CAD screening and offers a targeted intervention strategy for precision prevention at the interdisciplinary intersection of periodontology and cardiovascular medicine.