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

Run prediction

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)