Data-driven photocatalytic uranium extraction: machine learning insights into structure–activity relationships of metal–organic and covalent organic frameworks

Uranium extraction from seawater and uranium-containing wastewater represents a critical challenge for sustainable nuclear energy development. Photocatalysis has emerged as a promising green strategy due to its high efficiency, environmental compatibility, and potential for solar-driven operation. Among diverse photocatalysts, metal–organic frameworks (MOFs) and covalent organic frameworks (COFs) have shown exceptional potential owing to their tunable porosity, rich active sites, and adjustable electronic structures. This review comprehensively summarizes recent advances in MOF- and COF-based photocatalytic uranium extraction, emphasizing structure–activity relationships and strategies to enhance light absorption, charge separation, and stability. Particular attention is devoted to the integration of data-driven methods, especially machine learning (ML), to accelerate material discovery and performance prediction. The Random Forest algorithm is introduced as a representative approach for correlating structural descriptors—such as surface area, pore size, and band gap—with uranium removal efficiency. By bridging experimental data and predictive modeling, ML-guided photocatalysis offers a paradigm shift toward rational design of high-performance materials. Finally, the review highlights current challenges and future perspectives for developing intelligent, sustainable, and scalable photocatalytic systems for uranium recovery from complex aqueous environments.