Fig. 1

Summary of our computational approach: The MOF library was assembled by considering CoRE MOF and hMOF databases. Various structural, chemical, energy-based descriptors were generated for these MOFs and structural filtering resulted in 2,182 CoRE MOFs and 98,596 hMOFs. GCMC simulations were conducted to compute CO uptakes for 2,182 CoRE MOFs. Using this simulation data along with structural, chemical and energetic features of 2,182 CoRE MOFs, ML models were developed. These ML models predicted the CO uptakes of 98,596 hMOFs, followed by GCMC simulations for 2,884 hMOFs identified as potentially high-performing materials achieving CO uptakes larger than 0.8 mol/kg. The top-performing 20 CoRE MOFs and hMOFs were further analyzed for their linker and metal types.