Recent advances in computational modeling of MOFs: From molecular simulations to machine learning

The reticular chemistry of metal–organic frameworks (MOFs) allows for the generation of an almost boundless number of materials some of which can be a substitute for the traditionally used porous materials in various fields including gas storage and separation, catalysis, drug storage and delivery. The number of MOFs and their potential applications are growing so quickly that, when novel MOFs are synthesized, testing them for all possible applications is not practical. High-throughput computational screening approaches based on molecular simulations of materials have been widely used to investigate MOFs and identify the optimal MOFs for a specific application. Despite the growing computational resources, given the enormous MOF material space, computational identification of promising MOFs requires more efficient approaches in terms of time and effort. Leveraging data-driven science techniques can offer key benefits such as accelerated MOF design and discovery pathways via the establishment of machine learning (ML) models and interpretation of complex structure-performance relationships that can reach beyond expert intuition. In this review, we present key scientific breakthroughs that propelled computational modeling of MOFs and discuss the state-of-the-art approaches extending from molecular simulations to ML algorithms. Finally, we provide our perspective on the potential opportunities and challenges for the future of big data-driven MOF design and discovery.