MAPTrans: mutual attention transformer with dynamic meta-path pruning for drug repositioning

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning offers an innovative approach to drug discovery by identifying new therapeutic uses for existing drugs.
• Current computational methods struggle to integrate heterogeneous data and model multi-scale biological interactions effectively.

Purpose of the Study:

• To develop an advanced computational model for drug repositioning that addresses limitations in data integration and interaction modeling.
• To enhance the accuracy and efficiency of identifying potential drug-disease correlations.

Main Methods:

• Proposed MAPTrans, a novel method utilizing a multi-level meta-path aggregation strategy for dynamic drug and disease representation.
• Incorporated a multi-view importance assessment mechanism to optimize feature representation by filtering discriminating views.
• Designed a mutual attention mechanism Transformer architecture with cross-view interaction for multi-view information fusion.

Main Results:

• MAPTrans demonstrated superior performance compared to existing baseline models on multiple benchmark datasets.
• The method effectively integrates heterogeneous data and captures complex biological interactions.
• Optimized feature representation through multi-view assessment improved prediction accuracy.

Conclusions:

• MAPTrans represents a significant advancement in computational drug repositioning.
• The proposed model offers a more effective approach to drug discovery by leveraging multi-level and multi-scale biological insights.
• This method holds promise for accelerating the identification of novel therapeutic applications for existing drugs.