Multi-Task Learning in Analyzing the Working capacity of MOFs

The screening and rational design of materials with optimal methane capture performance remain key challenges for environmental and energy applications. Metal–organic frameworks (MOFs) are considered among the most promising candidates for methane capture. However, existing studies often emphasize single-point adsorption capacity, overlooking the dynamic mechanisms associated with pressure swing adsorption (PSA) required for practical applications. The construction of this dataset aims to address this challenge and provide insights into PSA-related mechanisms in MOFs.This dataset comprises geometric and chemical descriptors, as well as methane adsorption capacities under six different pressures, calculated for chemically reasonable MOF structures screened from eight publicly available datasets. In addition, we provide our in-house multitask learning (MTL) algorithm for predicting the gravimetric and volumetric working capacities of MOFs.<b>CIF files</b>: CIF files for 252,352 MOFs;<b>Geometric descriptors</b>: 14 geometric descriptors;<b>Chemical descriptors</b>: 176 chemical descriptors;<b>Methane_v, Methane_g</b>: Volumetric and gravimetric working capacities for methane adsorption, including methane adsorption data under six pressures across three application scenarios (landfill gas treatment, methane purification, and methane storage);<b>MTL4MOFsWC</b>: Python code for training the MTL models to predict the working capacity of methane adsorption in MOFs;<b>best_model_v_full, best_model_v_sim, best_model_g_full, best_model_g_sim</b>: Pre-trained MTL models.Please cite the following literature when you use this database:DOI: https://doi.org/10.1039/d5ta10538b