Using Machine Learning to Predict First-Order Reaction Rate Constants of PFAS Degradation

Per- and polyfluoroalkyl substances (PFAS) are environmentally persistent pollutants, posing challenges for effective remediation. This study presented a machine learning (ML) framework to predict the first-order reaction rate constant (k) of PFAS degradation across electrochemical, photochemical, and sonochemical processes. By integrating molecular descriptor (MD) and experimental conditions, the Extreme Gradient Boosting (XGB) achieved the best performance on test (R2 = 0.568, RMSE = 0.448) among Random Forest (RF), Extra Trees (ET), and Gradient Boosted Regression Trees (GBRT) models. Interpretability analysis revealed that experimental conditions, particularly reaction type and initial PFAS concentration, had greater influence on k than PFAS features. The model was further applied to predict the degradations of 2,631 PFAS from the OECD database, and clustering analysis identified structural groups with higher degradation potential, especially those containing weaker bond energies or labile functional groups, such as C–Br bond and -SO3H group. This study offers a scalable approach for assessing PFAS degradability.