Multi-View Fused Nonnegative Matrix Completion Methods for Drug-Target Interaction Prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Accurate drug-target interaction (DTI) prediction is vital for efficient drug discovery.
• Challenges include sparse interaction data and heterogeneous biological datasets.
• Existing methods often struggle with interpretability and scalability.

Purpose of the Study:

• To propose and validate novel multi-view fused nonnegative matrix completion methods for DTI prediction.
• To improve prediction accuracy, interpretability, and scalability by integrating heterogeneous similarity information.
• To address limitations of current DTI prediction approaches.

Main Methods:

• Developed two multi-view fused nonnegative matrix completion models.
• Integrated a nonnegative matrix completion framework with multi-graph Laplacian regularization.
• Employed a linear multi-view fusion mechanism with weights learned via linearly constrained quadratic programming.
• Utilized efficient proximal linearization-incorporated block coordinate descent algorithms for optimization.

Main Results:

• Models consistently outperformed state-of-the-art methods on four gold-standard and one real-world dataset.
• Ablation studies confirmed the significant contribution of each proposed model component.
• Scalability analysis demonstrated the computational efficiency of the developed approach.

Conclusions:

• The proposed multi-view fused nonnegative matrix completion methods offer a significant advancement in DTI prediction.
• These methods provide accurate, interpretable, and scalable solutions for drug discovery.
• The integration of heterogeneous data and advanced matrix completion techniques is effective.