Advanced drug-target interaction prediction using convolutional graph attention networks in expert systems.

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Predicting drug-target interactions (DTI) is essential for accelerating drug discovery and repurposing, reducing time and costs.
• Traditional experimental methods for DTI prediction are often time-consuming and expensive.

Purpose of the Study:

• To develop an advanced deep learning framework for accurate and efficient DTI prediction.
• To improve the reliability and efficiency of drug discovery and repurposing pipelines.

Main Methods:

• A novel deep learning framework, Convolutional Multilayer Extreme Adversarial Graph Attention-based Neural Network (CMEAG-ANN), was developed.
• A Fast Correlation-Based Gradient Naïve Bayes and Binary Pattern Selection (FC-GNBBPS) algorithm was used for robust feature extraction from DNA molecule-derived data.
• Graph attention algorithms integrated structural and evolutionary aspects of drugs and proteins using molecular fingerprints and PSSM-based annotations.

Main Results:

• The CMEAG-ANN model achieved high performance metrics: 99.17% accuracy, 99.11% precision, 98.83% recall, 98.96% F1-score, and 98.74% specificity.
• Experimental evaluations on benchmark datasets demonstrated the superiority of CMEAG-ANN over baseline models.
• Biologically grounded feature selection enhanced the model's predictive capabilities.

Conclusions:

• The proposed CMEAG-ANN framework significantly improves DTI prediction accuracy and efficiency.
• The study demonstrates the effectiveness of integrating graph-based neural networks with advanced feature selection for DTI prediction.
• This approach offers a more reliable and efficient method for drug discovery and repurposing.