Novel drug-target interactions via link prediction and network embedding

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Undiscovered drug-target interactions (DTIs) hinder drug discovery and repurposing.
• Existing computational methods struggle to integrate chemical and genomic data effectively.
• Lack of negative interaction data and target 3D structures limits DTI prediction accuracy.

Purpose of the Study:

• To develop an accurate computational framework for predicting drug-target interactions.
• To leverage graph embedding and machine learning for DTI prediction.
• To facilitate drug repurposing by identifying novel drug-target relationships.

Main Methods:

• DT2Vec pipeline utilizes graph embedding to map drug-drug and protein-protein similarity networks.
• Drug and target embedding vectors are concatenated as input features for binary classification.
• The model was trained and validated on ChEMBL repository data with experimentally validated interactions.

Main Results:

• DT2Vec achieved competitive results compared to existing graph similarity-based algorithms.
• The model successfully predicted novel drug-target interactions from large datasets.
• Molecular docking was used to evaluate the credibility of predicted novel DTIs.

Conclusions:

• DT2Vec effectively integrates chemical and genomic spaces into low-dimensional vectors.
• The method shows promise for predicting novel drug-target interactions.
• DT2Vec serves as a valuable tool for accelerating drug discovery and repurposing efforts.