Guest–host interactions in confined environments are crucial for many chemical and physical processes, yet their quantitative representation in computations remains rarely reported. Here, we introduce a physically interpretable interaction in confined environment (ICE) descriptor to characterize guest–host interactions in metal–organic frameworks (MOFs). The descriptor is constructed by first identifying accessible sites through Voronoi analysis of the host framework, followed by the vectorial encoding of local interaction forces. The results reveal distinct mechanistic sensitivities: adsorption correlates with the integrated strength of interaction forces, whereas diffusion is more sensitive to the spatial distribution of the sites. Owing to its physical grounding and transferability, ICE provides a robust framework for modeling interaction-driven phenomena in confined environments and advancing data-driven material design.
Wang et al. (Mon,) studied this question.