Drug-target interaction prediction based on metapaths and simplified neighbor aggregation.

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Drug-target interaction (DTI) prediction is vital for drug repositioning and discovery.
• Current metapath-based methods use attention mechanisms, which can be unstable with noisy, imbalanced biological network data.
• Attention mechanisms risk overfitting and reduced efficiency in small-scale datasets.

Purpose of the Study:

• To propose a simplified mean aggregation method for DTI prediction.
• To address the limitations of attention mechanisms in noisy biological networks.
• To enhance the stability, efficiency, and accuracy of DTI prediction models.

Main Methods:

• Implemented a simplified mean aggregation approach for DTI prediction.
• Reduced model complexity and noise interference through average aggregation.
• Evaluated the method on three heterogeneous biological datasets.

Main Results:

• SNADTI outperformed 12 leading DTI prediction methods across two metrics.
• Demonstrated significant reduction in training time and enhanced computational speed.
• Achieved superior prediction accuracy, stability, and reproducibility.

Conclusions:

• SNADTI offers a practical and effective solution for DTI prediction in biological networks.
• The average aggregation method mitigates noise and overfitting issues common in DTI prediction.
• SNADTI presents a computationally efficient and reliable alternative for drug discovery research.