Digital twin modeling of root canal dentin solubility: A pilot study using machine learning analysis

Abstract Introduction: Dentin solubility influences the durability of endodontic treatment, as excessive demineralization may compromise long-term tooth stability. Trace element incorporation has been shown to alter dentin mineral properties, but systematic evaluation of their effects remains limited. This study aimed to develop a digital twin model to predict dentin solubility in response to trace element incorporation to optimize endodontic biomaterials. Materials and Methods: Thirty dentin specimens were treated with trace elements, and the solubility was measured under controlled conditions. The five elemental concentration predictors were standardized using StandardScaler. A Random Forest regression model was trained on 80% of the data and evaluated on the remaining 20% as a holdout test set, with performance assessed using mean squared error (MSE) and coefficient of determination metrics. Feature importance analysis identified the most influential elements. Results: The model achieved strong predictive performance ( R 2 = 0.74; MSE = 0.031). Feature importance analysis identified strontium and barium as the most influential predictors of solubility (26% and 23% of model importance), and produced the greatest reduction in solubility (6.00% ±0.16% and 6.20% ±0.16%, respectively) compared to control (7.10% ±0.16%). One-way analysis of variance confirmed statistically significant differences among groups ( P < 0.0001), and Tukey post hoc testing validated that all trace element groups differed significantly from control ( P < 0.001). Conclusions: Within the limits of this pilot study, the digital twin model was able to predict dentin solubility based on trace element composition and to identify elements associated with reduced solubility. These findings support the feasibility of using data-driven modeling as a screening tool to guide future studies in endodontic biomaterial development.