Authors: Sartaaj Takrim Khan, Seyed Mohamad Moosavi

This repository contains data that was used in the study "Connecting metal-organic framework synthesis to applications with multimodal machine learning". 
For demonstrations on the code (deep learning models, preprocessing, etc.), please visit https://github.com/AI4ChemS/XRayPro/. 

####################################################################################################
1. weights/

This file contains the weights for the following properties:
- CoRE-MOF (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- BW20K (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- ARABG (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- QMOF (band gap)
- hMOF (H2 capacity for pressure swing between 100 to 1 bar at 77K, Xe uptake at 10 bar)

####################################################################################################
2. pxrd/

This file contains PXRD patterns (in .npy arrays) for:
- CoRE-MOF
- BW20K
- ARABG 
- QMOF
- hMOF

To load these, use the following code: arr = np.load('pxrd/database_of_interest/cif_id.npy'). These are computed from pymatgen and is a nested array with the format:
[[theta], [intensity]]

####################################################################################################
3. precursors/

This file contains the precursors (in .txt) for:
- CoRE-MOF
- BW20K
- ARABG
- QMOF
- hMOF

Each precursor is in the format [metal_type].[SMILES of organic_linker].

####################################################################################################
4. labels/

This file contains the csvs for the labels used in this study (alongside recommendations using thresholds) for:
- CoRE-MOF (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- BW20K (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- ARABG (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- QMOF (band gap)
- hMOF (H2 capacity for pressure swing between 100 to 1 bar at 77K, Xe uptake at 10 bar)

####################################################################################################
5. figures/

This figure contains the code and the csvs required to generate Figures 2, 3 and 5 from the main text. The folder names are as follows:
- fig2
- fig3
- fig5

####################################################################################################
6. splits/

This folder contains sets of random splits that correspond to the weights folder uploaded. This has been made for the user to use in predictions.ipynb as a demonstration on making predictions on properties, given CIFs.

- CoRE-MOF (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- BW20K (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- ARABG (CH4 uptake at 65 bar, CO2 uptake at 0.15 bar, largest_included_sphere (Di), density, logKH_CO2)
- QMOF (band gap)
- hMOF (H2 capacity for pressure swing between 100 to 1 bar at 77K, Xe uptake at 10 bar)