LM-DTI: a tool of predicting drug-target interactions using the node2vec and network path score methods.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Drug-target interaction (DTI) prediction is crucial for discovering drug functions and repositioning existing drugs.
• Large-scale biological networks offer opportunities for identifying drug-related target genes, driving computational DTI prediction methods.
• Conventional computational methods have limitations that necessitate novel approaches.

Purpose of the Study:

• To propose LM-DTI, a novel computational tool for DTI prediction.
• To integrate lncRNA and miRNA information for enhanced DTI prediction accuracy.
• To develop a scalable and efficient tool for drug repositioning.

Main Methods:

• Constructed a heterogeneous information network with drug, target, lncRNA, and miRNA nodes.
• Employed graph embedding (node2vec) to obtain feature vectors for drug and target nodes.
• Calculated network path scores using DASPfind, merged features, and used XGBoost for prediction.

Main Results:

• LM-DTI achieved a high prediction performance with an AUPR of 0.96.
• Demonstrated significant improvement over conventional DTI prediction tools.
• Validated LM-DTI's effectiveness through literature and database searches, confirming its scalability and efficiency.

Conclusions:

• LM-DTI represents a powerful and efficient tool for drug repositioning.
• The integration of lncRNA and miRNA data enhances DTI prediction accuracy.
• LM-DTI is freely accessible, facilitating broader application in drug discovery.