This archive contains the 1M1L3D dataset and machine learning models implemented as a Jupyter Notebook described in Angewandte Chemie Communication "Machine Learning Prediction of Metal-Organic Framework Guest Accessibility from Linker and Metal Chemistry" by Rémi Pétuya, Samantha Durdy, Dmytro Antypov, Michael W. Gaultois, Neil G. Berry, George R. Darling, Alexandros P. Katsoulidis, Matthew S. Dyer, and Matthew J. Rosseinsky

https://doi.org/10.1002/anie.202114573

Please cite the corresponding paper if you used the 1M1L3D dataset or these machine learning models in your work.

The Jupyter Notebook 

MOF_guest_accessibility_ML_sequence_models.ipynb

hereafter presents the entire training procedure of the sequence of classification of 3D MOF porosity range reported in the paper. The final section of this Notebook allows the user to input their own combination of linker and metal for sequential prediction of the porosity range.

This notebook uses the following four input files in the CSV format:

1M1L3D_summary.csv - the list of metal and linker constituents of 14,296 reported 3D MOF structures

1M1L3D_metal_descriptors.csv - the list of 6 metal features, as listed in elemental_descriptors.csv arranged for all MOF structures in the 1M1L3D dataset

1M1L3D_Mordred_2D_descriptors.csv - the list of 1,613 linker features for all MOF structures

elemental_descriptors.csv - the list of 6 elemental features for 85 elements found in the 1M1L3D dataset