DrugTar-Protein Druggability Prediction

This repository contains the Python code and dataset for the paper titled “DrugTar improves druggability prediction by integrating large language models and gene ontologies” by Niloofar Borhani, Iman Izadi, Ali Motahharynia, Mahsa Sheikholeslami, and Yousof Gheisari. The paper is available at https://doi.org/10.1093/bioinformatics/btaf360.

The online tool is available at www.drugtar.com.

Feature Generation

• esm2_embedder.py: Generates ESM-2 embeddings for each protein sequence.
• feats_extraction.py: Extracts features (ESM-2 embeddings and Gene Ontology (GO) terms) and corresponding labels, creating feature matrices as DataFrames.

Training and Evaluation with k-Fold Cross Validation

To train the DrugTar model and validate it using k-fold cross-validation, run the following command:

python scripts/main.py --phase='train' \
--batch_size=32 --num_epochs=110 --k_fold=10 --init_lr=0.0002 \
--save_model='True' --DNN_dims='128_64_32' --dropout=0.5 \
--feat_selected=4000 \
--train_feature='feature/train_feature.pkl' \
--train_label='feature/train_label.pkl' \
--test_feature='feature/test_feature.pkl' \
--test_label='feature/test_label.pkl' \
--model_file='models/' \
--save_pred_file='prediction/druggability_scores.csv'

Testing the Trained Model

To test the performance of the trained DrugTar model on independent test data:

python scripts/main.py --phase='test' \
--batch_size=32 --num_epochs=110 --init_lr=0.0002 \
--save_model='True' --DNN_dims='128_64_32' --dropout=0.5 \
--feat_selected=4000 \
--test_feature='feature/test_feature.pkl' \
--test_label='feature/test_label.pkl' \
--model_file='models/' \
--save_pred_file='prediction/druggability_scores.csv'

Prediction for Unlabeled Proteins

To predict druggability scores for unlabeled proteins using the trained DrugTar model:

python scripts/main.py --phase='prediction' \
--batch_size=32 --num_epochs=110 --init_lr=0.0002 \
--save_model='True' --DNN_dims='128_64_32' --dropout=0.5 \
--feat_selected=4000 \
--test_feature='feature/test_feature.pkl' \
--model_file='models/' \
--save_pred_file='prediction/druggability_scores.csv'

Contact Information

For further inquiries, please contact.

Niloofar Borhani
Ph.D. Student, Control Engineering
Isfahan University of Technology
Email: n.borhani@ec.iut.ac.ir
CV: Google Scholar