Tripartite Network-Based Repurposing Method Using Deep Learning to Compute Similarities for Drug-Target Prediction.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug discovery is complex and resource-intensive, driving interest in drug repositioning.
• Similarity-based methods are widely used for predicting drug-target associations.
• Integrating heterogeneous network topology with deep learning offers potential for improved drug-target prediction.

Purpose of the Study:

• To develop a novel similarity-based drug-target prediction method.
• To leverage deep learning and heterogeneous network topology for enhanced prediction accuracy.
• To improve the flexibility and effectiveness of drug repositioning strategies.

Main Methods:

• Utilized DeepWalk, a deep learning algorithm, to compute vertex similarities within a Linked Tripartite Network (LTN).
• Constructed LTN as a heterogeneous network integrating diverse biomedical datasets.
• Employed drug-based similarity inference (DBSI) and target-based similarity inference (TBSI) for predicting drug-target associations.

Main Results:

• The proposed method effectively utilizes deep learning and heterogeneous network topology for drug-target prediction.
• Achieved more promising results compared to existing topology-based similarity computation methods in prior experiments.
• Demonstrated the potential of integrating network topology with deep learning for drug repositioning.

Conclusions:

• The developed method offers a more flexible and accurate approach to drug-target prediction.
• Deep learning combined with heterogeneous network topology significantly enhances similarity-based prediction.
• This approach holds promise for accelerating drug repositioning and drug discovery efforts.