README.md

Data for “Discovery of Crystallizable Organic Semiconductors with Machine Learning”

install required packages

pip install ./requirements.txt

Note: You have to have a licensed and installed version of AlvaDesc software to calculate the descriptors

File’s description

├── data
│   ├── Dataset_dHm.csv
│   └── Dataset_Tm.csv
├── descriptors
│   ├── SITable_Descriptors_dHm_list.csv
│   └── SITable_Descriptors_Tm_list.csv
├── models
│   ├── Model_dHm.pkl
│   └── Model_Tm.pkl
├── Predict.py
├── README.md
├── requirements.txt
├── run_Predict.sh
└── test
    ├── Test_mols.csv
    └── Test_mols_pred.csv

The training sets for dHm and Tm are in Dataset_dHm.csv and Dataset_Tm.csv. The pretrained models serialized in pickle format are Model_dHm.pkl and Model_Tm.pkl.

The script Predict.py runs one of those models, calculates the set of appropriate descriptors specified in SITable_Descriptors_dHm_list.csv (or SITable_Descriptors_Tm_list.csv) and makes predictions for a set of input molecules.

An example of input file is Test_mols.csv, an example of the output is Test_mols_pred.csv

Bash script with an example of specification of input parameters is provided -- run_Predict.sh.

If you use this data, please cite the following work:

Discovery of Crystallizable Organic Semiconductors with Machine Learning Holly M. Johnson, Filipp Gusev, Jordan T. Dull, Yejoon Seo, Rodney D. Priestley, Olexandr Isayev, and Barry P. Rand Journal of the American Chemical Society 2024 146 (31), 21583-21590 DOI: 10.1021/jacs.4c05245