Three classification models are provided in this repository:
- reaction-EC classification model
- substrate-EC classification model
- product-EC classification model
- numpy 1.19.2
- sklearn 0.23.2
- pandas 1.1.3
The molecules of reactions/substrates/products were embedded using mol2vec (https://github.com/samoturk/mol2vec).
The EC numbers were embedded using EC2Vec (https://github.com/MengLiu90/EC2Vec).
The embeddings were concatenated together as the input features to the classifiers.
Please refer to the data in the ./Data/ directory for examples.
To execute the classifiers, prepare the data as per the provided instructions and place it within the ./Data/ directory.
The Python scripts RF_product_EC_classifier.py, RF_reaction_EC_classifier.py, and RF_substrate_EC_classifier.py implement respective classifiers using the train/test split protocol.
The Python scripts RF_product_EC_classifier_5_fold_cv.py, RF_reaction_EC_classifier_5_fold_cv.py, and RF_substrate_EC_classifier_5_fold_cv.py implement respective classifiers using the 5-fold cross-validation protocol.
Run RF_product_EC_classifier_5_fold_cv.py to obtain performance metrics for predicting the existence of a product-EC pair.
Run RF_reaction_EC_classifier_5_fold_cv.py to obtain performance metrics for predicting the existence of a reaction-EC pair.
Run RF_substrate_EC_classifier_5_fold_cv.py to obtain performance metrics for predicting the existence of a substrate-EC pair.
As the compounds in the data we used to train the classifier may differ significantly from those in your dataset, and it is simple and efficient to train a Ramdom Forest model, you can directly train your own model with your data and the tained model will be automatically saved in the ./Trained_model/ directory.