MPGraph: multi-view penalised graph clustering for predicting drug-target interactions.

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Experimental determination of drug-target interactions is costly and time-consuming.
• Computational methods are essential for efficient prediction of these interactions.
• Existing methods may not fully leverage diverse data types for interaction prediction.

Purpose of the Study:

• To develop advanced computational models for predicting drug-target interactions.
• To integrate multiple data views (structural, chemical, gene expression) for enhanced prediction accuracy.
• To identify novel drug-target interactions for drug repositioning.

Main Methods:

• Proposed a single-view penalised graph (SPGraph) clustering approach integrating drug structure and protein sequence data.
• Applied SPGraph to chemical data (drug response, gene expression) from NCI-60 cell lines.
• Generalized SPGraph to a multi-view penalised graph (MPGraph) model for integrating structural and chemical views.

Main Results:

• SPGraph demonstrated improved prediction accuracy on a smaller scale.
• MPGraph achieved approximately 10% improvement in prediction accuracy compared to baseline methods.
• Identified potential new targets for 22 FDA-approved drugs, with some findings supported by existing literature.

Conclusions:

• The developed SPGraph and MPGraph models offer effective computational strategies for predicting drug-target interactions.
• MPGraph provides a significant advancement by integrating multi-view data, leading to higher prediction accuracy.
• The identified drug-target interactions hold promise for accelerating drug repositioning and discovery efforts.