Multi-domain based heterogeneous network for drug-target interaction prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Computational approaches are vital for predicting drug-target interactions (DTIs) in drug discovery.
• Existing methods often fail to capture substructural details of drug-target pairs (DTPs) and generalize poorly to new data.

Purpose of the Study:

• To develop an interpretable and generalizable framework for DTI prediction.
• To address limitations in molecular representation and dataset domain specificity in current DTI prediction models.

Main Methods:

• Introduced MHNF-DTI, a Multi-domain based Heterogeneous Network Framework.
• Utilized a transformer encoder integrated with multilayer graph attention networks to capture DTP sub-structural properties.
• Constructed a multi-domain label reversal dataset to enhance model generalization and mitigate ligand bias.

Main Results:

• MHNF-DTI effectively captures sub-structural properties of drugs and targets.
• The framework demonstrates improved interpretability and generalization capabilities.
• Experimental results show superior DTI prediction performance compared to state-of-the-art methods.

Conclusions:

• MHNF-DTI offers a robust solution for accurate and interpretable DTI prediction.
• The proposed framework advances computational approaches in drug discovery and chemogenomics.
• MHNF-DTI shows significant potential for identifying novel drug-target interactions.