Deep Learning-Based Modeling of Drug-Target Interaction Prediction Incorporating Binding Site Information of...

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Chemogenomics, or proteochemometrics, utilizes computational methods to predict drug-target interactions, crucial for early drug discovery and identifying off-target effects.
• Accurate prediction of ligand-target interactions aids in identifying potential therapeutic candidates and understanding their safety profiles.

Purpose of the Study:

• To develop a computationally efficient deep learning model for predicting unknown ligand-target interactions.
• To utilize one-dimensional SMILES for ligands and binding site residues for proteins as input features.
• To compare the performance of the proposed deep learning model against traditional machine learning methods.

Main Methods:

• A Convolutional Neural Network (CNN) deep learning model, termed DeepPS, was developed using one-dimensional SMILES strings for drugs and motif-rich binding pocket subsequences for proteins.
• The model was trained and evaluated using expert-based features and compared against shallow feature-based machine learning approaches.
• Performance was assessed using Mean Squared Error (MSE) and Area Under the Precision-Recall Curve (AUPR) metrics.

Main Results:

• The proposed DeepPS model demonstrated comparable or superior performance on MSE and AUPR metrics compared to shallow machine learning methods.
• DeepPS proved to be computationally more efficient than deep learning models trained on full-length protein sequences.
• The study validates the effectiveness of using simplified molecular and protein sequence representations for predicting drug-target interactions.

Conclusions:

• Integrating protein structural information into drug-target interaction prediction models can enhance interpretability, throughput, and applicability for large datasets.
• The DeepPS model offers a promising, efficient computational strategy for early-stage drug discovery and development.
• Further research incorporating structural data could significantly advance the field of chemogenomics.