High-Throughput Virtual Screening of Biometal-Organic Frameworks for O₂/N₂ Separation

The separation/purification of oxygen (O-2) from air is of great significance in the biomedical field. Biometal-organic frameworks (bio-MOFs), as a class of promising alternatives to traditional adsorbents, have attracted widespread interest. This paper proposes a strategy for screening high-performance bio-MOFs based on machine learning (ML) and molecular simulation methods. First, nontoxic and cost-effective bio-MOFs, namely, desired bio-MOFs, are selected from MOF databases using the binary decision tree method. Next, 15 descriptors, including nine structural descriptors and six chemical descriptors, are calculated to characterize the desired bio-MOFs. Next, the random forest (RF) algorithm is adopted to map the relationship between descriptors and the target property, where target properties are calculated by the grand canonical Monte Carlo (GCMC) results. High-throughput screening of the high-performance desired bio-MOFs is performed using the established RF model. Finally, high-performance desired bio-MOFs are obtained for O-2/N-2 adsorption separation, and their structure-property relationships are also analyzed.