Solv_GNN_SSD

Python environment setup

The 'tf24gpu.yml' file contains the information about dependencies required to run model training and prediction.

conda env create -f tf24gpu.yml

conda activate tf24gpu

First, create 'molecules_to_predict.csv' file that contains the following columns:

Each row: solute-solvent pairs whose Gibbs free energy of solvation needs to be predicted
'can_smiles_solute' column: SMILES string of solute
'can_smiles_solvent' column: SMILES string of solvent
'DGsolv' column: Experimental Gibbs free energy of solvation of a given solute-solvent pair (if available, if there are no experimental values, this column can be left blank)

in the directory where 'main.py' is located. Next, run

python main.py -predict_df -modelname SSD_models/student35  # any models in the 'model_files' folder can be read for the prediction.

Then, a user can find 'molecules_to_predict_results.csv' file which contains the predicted Gibbs free energies of solvation. ('predicted' column)

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
Fig1A+1B		Fig1A+1B
Fig5_stratified_sampling		Fig5_stratified_sampling
Fig7_tsne		Fig7_tsne
Figs3+4		Figs3+4
Table1+QM_DB		Table1+QM_DB
__pycache__		__pycache__
data		data
json_with_atom_formal_charges		json_with_atom_formal_charges
model_files		model_files
shap		shap
two_appl_examples		two_appl_examples
Fig1B_plot.ipynb		Fig1B_plot.ipynb
Fig6.ipynb		Fig6.ipynb
README.md		README.md
expdb_combisolvqm_overlapped.csv		expdb_combisolvqm_overlapped.csv
gnn.py		gnn.py
main.py		main.py
tf24gpu.yml		tf24gpu.yml