TraNce: Type-aware hypergraph neural network with biological mediators for drug repositioning

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Computational drug repositioning (DR) accelerates pharmaceutical discovery by identifying new uses for existing drugs.
• Hypergraph-based DR approaches capture complex biological interactions but often neglect diverse biological mediators.
• Conventional hypergraph neural networks (HGNNs) struggle with type-agnostic aggregation, leading to semantic confusion.

Purpose of the Study:

• To propose a novel type-aware HGNN model, TraNce, for improved computational drug repositioning.
• To incorporate diverse biological entities (genes, phenotypes, pathways) into hypergraph structures.
• To address the type-agnostic aggregation challenge in HGNNs for enhanced semantic boundary preservation.

Main Methods:

• Constructed hypergraphs by integrating drugs, diseases, and bridging biological intermediates (genes, phenotypes, pathways) from biomedical knowledge graphs.
• Developed type-aware message propagation and aggregation schemes within the HGNN framework.
• Implemented an adaptive element-wise gating mechanism for integrating type-specific message embeddings.

Main Results:

• TraNce demonstrated significant effectiveness across three large-scale datasets.
• The model achieved notable performance gains, with up to 2.17% AUC and 2.69% AUPR improvements over baseline methods.
• Experimental validation confirmed the model's ability to handle heterogeneous biological entity types.

Conclusions:

• TraNce offers a powerful new approach for computational drug repositioning by effectively leveraging diverse biological data.
• The type-aware methodology overcomes limitations of conventional HGNNs, leading to more accurate predictions.
• The proposed model holds promise for accelerating pharmaceutical discovery and identifying novel therapeutic indications.