Optimizing a Diagnostic Model of Periodontitis by Using Targeted Proteomics

Periodontitis (PD), a widespreadchronic infectious disease,compromisesoral health and is associated with various systemic conditions andhematological alterations. Yet, to date, it is not clear whether serumprotein profiling improves the assessment of PD. We collected generalhealth data, performed dental examinations, and generated serum proteinprofiles using novel Proximity Extension Assay technology for 654participants of the Bialystok PLUS study. To evaluate the incrementalbenefit of proteomics, we constructed two logistic regression modelsassessing the risk of having PD according to the CDC/AAP definition;the first one contained established PD predictors, and in addition,the second one was enhanced by extensive protein information. We thencompared both models in terms of overall fit, discrimination, andcalibration. For internal model validation, we performed bootstrapresampling (n = 2000). We identified 14 proteins,which improved the global fit and discrimination of a model of establishedPD risk factors, while maintaining reasonable calibration (area underthe curve 0.82 vs 0.86; P < 0.001). Our resultssuggest that proteomic technologies offer an interesting advancementin the goal of finding easy-to-use and scalable diagnostic applicationsfor PD that do not require direct examination of the periodontium.