Program Description

This program consists of three main Python scripts for data preprocessing, model training, and generating SHAP explanations.

The required Python packages are

• numpy
• pickle
• json
• rdkit
• pandas
• scikit-learn (sklearn)
• xgboost
• imbalanced-learn (imblearn)
• shap

1. compute_fingerprints.py

This script computes fingerprints from molecular structures specified in a DataFrame.

Usage:

python compute_fingerprints.py -dfp dataframe.csv -ext ext -it smile -sc Smiles -tc AT_down_05
python compute_fingerprints.py -dfp dataframe.csv -ext ext -it mol -sc Structure -tc AT_down_05

Parameters:

• -dfp, --dataframe_path: Path to the DataFrame CSV file.
• -ext, --extension_name: Experiment extension name.
• -it, --input_type: Input type, either 'smile' or 'mol'.
• -sc, --structure_column: Name of the column containing molecular structures.
• -tc, --target_column: Name of the target column.

2. train_models.py

This script trains machine learning models using specified fingerprints and parameters.

Usage:

python train_models.py -ext ext
python train_models.py -ext ext -s f1 recall
python train_models.py -ext ext -fp ap dl tt
python train_models.py -ext ext -fp all_count
python train_models.py -ext ext -m rf xgb
python train_models.py -ext ext -ni 10
python train_models.py -ext ext -sp oversampling
python train_models.py -ext ext -ns 100 -fs chi2
python train_models.py -ext ext -fp all_count

Parameters:

• -fp, --fingerprints: Chosen fingerprints.
• -m, --models: Selected models.
• -s, --scores: Cross-validation scores.
• -ext, --extension_name: Experiment extension name.
• -tg, --target: Target name.
• -rs, --random_seed: Random seed.
• -sp, --sampling: Over/undersampling method.
• -ni, --n_iter: Number of iterations.
• -fs, --feature_sel: Feature selection method.
• -ns, --n_sel: Number of selected features.

3. compute_shap.py

This script computes SHAP explanations for model predictions.

Usage:

python compute_shap.py -ext ext
python compute_shap.py -ext ext -s f1 recall
python compute_shap.py -ext ext -fp ap dl tt
python compute_shap.py -ext ext -fp all_count
python compute_shap.py -ext ext -m rf xgb
python compute_shap.py -ext ext -ni 10
python compute_shap.py -ext ext -sp oversampling
python compute_shap.py -ext ext -ns 100 -fs chi2
python compute_shap.py -ext ext -fp all_count

Parameters:

• -fp, --fingerprints: Chosen fingerprints.
• -m, --models: Selected models.
• -s, --scores: Cross-validation scores.
• -ext, --extension_name: Experiment extension name.
• -rs, --random_seed: Random seed.
• -sp, --sampling: Over/undersampling method.
• -ni, --n_iter: Number of iterations.
• -fs, --feature_sel: Feature selection method.
• -ns, --n_sel: Number of selected features.

---