Advancing metal organic frameworks screening and remediation condition optimization for removing persistent organic...

Metal organic frameworks (MOF) have become increasingly important for removing persistent organic pollutants (POPs). However, achieving precise and rational design of MOF based composites, as well as optimizing operational conditions for real applications, remains challenging. Many studies still rely on empirical trial and error approaches, which limits efficiency and scalability. In this work, we establish an integrated framework that combines machine learning with theory calculations to predict, interpret, and optimize POPs removal by MOF composites. The construted database includes physicochemical descriptors, operational parameters, and additional aromatic related features that contribute to capture the complexity of MOF and pollutant interactions. Several conventional and advanced learning algorithms were evaluated, the XGBoost optimized through the covariance matrix adaptation evolution strategy achieved the best balance of predictive accuracy, robustness, and interpretability. The machine learning model also identified π-π interaction as major contributors to adsorption behavior. SHAP and partial dependence analyses further indicated that starting concentration, specific surface area, and particle scale strongly influence removal behavior. These analyses also revealed the importance of water chemistry factors, which offers practical guidance for adjusting treatment conditions. The complement theory calculations show that increasing aromatic ring content and adjusting offset parallel stacking enlarge the interaction contact area, strengthen electrostatic complementarity, and enhance orbital overlap. Overall, this combined data driven and quantum level method links macroscopic performance prediction with microscopic interaction mechanisms. It provides a practical strategy for guiding MOF structural design and optimizing treatment conditions to achieve efficient and sustainable POPs remediation.