A heterogeneous network embedding framework for predicting similarity-based drug-target interactions.

Area of Science:

• Computational biology
• Bioinformatics
• Network science

Background:

• Drug-target interactions (DTIs) are crucial for drug development.
• Current DTI prediction methods often overlook multi-network features.
• Integrating diverse biological data is essential for accurate DTI prediction.

Purpose of the Study:

• To propose a novel Network EmbeDding framework in mulTiPlex networks (NEDTP) for DTI prediction.
• To effectively extract and merge drug and target features from heterogeneous information networks.
• To address limitations in existing DTI prediction methodologies.

Main Methods:

• Constructed a similarity network using 15 heterogeneous information networks.
• Employed random walks to extract node topology information and generate low-dimensional vectors.
• Utilized a Gradient Boosting Decision Tree model for DTI classification.

Main Results:

• NEDTP demonstrated superior accuracy in DTI prediction compared to state-of-the-art algorithms.
• Validated new DTI predictions through multiple experimental perspectives.
• Achieved significant improvements in predicting drug-target interactions.

Conclusions:

• NEDTP offers an effective approach for DTI prediction by leveraging multi-network information.
• The framework provides a robust solution for negative sampling in DTI prediction.
• This study contributes novel insights and methods to the field of computational drug discovery.