It looks like a great use case for machine learning, as the author states:
“The number of MOF structures is massive—there are thousands of experimentally synthesized structures, but also many more hypothesized ones—creating a need for efficient tools and approaches to quickly identify MOFs best suited for a given applications.”