Drug-target interaction prediction by integrating multiview network data.

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Drug-target interaction (DTI) prediction is crucial for drug repositioning, discovery, and design.
• Existing DTI prediction methods often rely on single-view data, limiting prediction stability and accuracy.
• High-throughput technologies generate extensive datasets, necessitating advanced computational algorithms for DTI identification.

Purpose of the Study:

• To develop a more stable and accurate multiview DTI prediction method by integrating drug and target data from different perspectives.
• To propose a clustering-based approach that leverages network consistency across multiple data views.
• To identify novel potential drug-target pairs for further research.

Main Methods:

• A single-view DTI prediction model formulated as an optimization problem to identify clusters in drug and target similarity networks.
• Extension of the model to multiview network data by maximizing cluster consistency across different views.
• Development of an approximation method to solve the formulated optimization problem for efficient computation.

Main Results:

• The proposed multiview DTI prediction algorithm demonstrated superior accuracy compared to existing single-view methods.
• Application to a two-view dataset resulted in the prediction of 54 potential drug-target interactions.
• Analysis of predicted DTIs, including similarity, enrichment, and gene analysis, indicated a high probability of their validity.

Conclusions:

• The multiview DTI prediction method effectively integrates diverse data sources for improved prediction accuracy.
• The clustering-based approach offers a robust framework for identifying reliable drug-target pairs.
• The predicted DTIs warrant further experimental validation, potentially accelerating drug discovery and development pipelines.