DTI-HeNE: a novel method for drug-target interaction prediction based on heterogeneous network embedding

Area of Science:

• Bioinformatics
• Computational Drug Discovery
• Network Science

Background:

• Drug-target interaction (DTI) prediction is crucial for efficient drug repurposing.
• Machine learning, particularly graph embedding, is increasingly used for DTI prediction.
• Existing graph embedding methods struggle to fully utilize heterogeneous information in drug-target networks.

Purpose of the Study:

• To develop a specialized DTI prediction method that effectively utilizes heterogeneous network information.
• To improve the quality of embeddings for drug-target pairs to enhance DTI prediction accuracy.

Main Methods:

• Propose DTI-HeNE (DTI based on Heterogeneous Network Embedding), a method designed for bipartite DTI networks.
• Decompose the heterogeneous DTI network into bipartite DTI, drug homogeneous, and target homogeneous networks.
• Extract features from sub-networks, integrate them using pathway information, and generate embedding vectors.
• Utilize a random forest model for novel DTI prediction.

Main Results:

• DTI-HeNE effectively generates high-quality embeddings for drug-target pairs.
• The method successfully utilizes both bipartite DTI relations and auxiliary similarity information.
• Experimental results demonstrate the method's capability in discovering novel DTIs.

Conclusions:

• The proposed DTI-HeNE method significantly improves feature learning for heterogeneous drug-target interaction networks.
• This approach enhances the discovery of novel drug-target interactions.