A graph neural network-based interpretable platform for small molecule pKa prediction.
input\: Input data, including the training dataset, external validation set, and ionizable site file.model\: The GraFpKa model.results\:Used to store prediction results.temp\: A temporary folder for storing temporary data.visualization\: Visualization results, used to store the results of interpretability analysis.
-
The project provides an environment file env_GraFpKa.yml. You can install the environment by running the following command:
conda env create -f env_GraFpKa.yml -
After that, activate the environment with:
conda activate env_GraFpKa
- If you need to train the model:
- You can run the following command for pretraining:
python pretraining.py - Then, run the following command for fine-tuning:
python fine_tuning.py
- You can run the following command for pretraining:
python predicting.py [--s SMILES...] [--i INPUT_FILE] [--o OUTPUT_FILE]
--s SMILES...: Input multiple SMILES strings directly via the command line. (Mutually exclusive with --i).
--i INPUT_FILE: Provide a CSV file containing SMILES strings (one SMILES per line, no header). (Mutually exclusive with --s).
--o OUTPUT_FILE: Optional. Specify the output CSV file path for saving pKa predictions. If not provided, the default output path is results/predictions.csv.
Predicting pKa values with SMILES input directly:
python predicting.py --s "CC(=O)OC1=CC=CC=C1C(=O)O" "Cc1ccccc1NN=C(C#N)C#N" "COc1cccc(S(N)(=O)=O)c1"
Predicting pKa values with a CSV input file:
python predicting.py --i input/input_example.csv --o results/predictions.csv
python interpretability.py [--s SMILES...] [--i INPUT_FILE] [--o OUTPUT_DIR] [--t THRESHOLD]
--s SMILES...: Input multiple SMILES strings directly via the command line. (Mutually exclusive with --i).
--i INPUT_FILE: Provide a CSV file containing SMILES strings (one SMILES per line, no header). (Mutually exclusive with --s).
--o OUTPUT_FILE: Optional. Specify the output directory to save interpretability analysis images. The directory must end with /. If not provided, the default output path is visualization/.
--t THRESHOLD: Optional.Specify the threshold for the color bar in the interpretability analysis. Default is -1 for automatic thresholding.
Performing interpretability analysis with SMILES input directly:
python interpretability.py --s "CC(=O)OC1=CC=CC=C1C(=O)O" "Cc1ccccc1NN=C(C#N)C#N" "COc1cccc(S(N)(=O)=O)c1" --t -1
Performing interpretability analysis with a CSV input file:
python interpretability.py --i input/input_example.csv --o visualization/