GraphRepur

GraphRepur predicts candidate drugs for breast cancer by integrating drug network-based and drug signature-based analyses with a GraphSAGE-based graph neural network.

Key Features:

• Method integration: Integrates drug network-based approaches and drug signature-based analyses to combine network relationships and drug-induced gene expression signatures.
• Graph model: Implements a GraphSAGE-based graph neural network to learn representations from drug networks.
• Differential expression: Uses differentially expressed genes associated with breast cancer as biological input signals.
• Drug-exposure data: Incorporates drug-exposure gene expression data to capture drug-specific transcriptional signatures.
• Drug-drug links: Utilizes drug-drug link information to represent topological relationships between drugs within the network.
• Signature and topology fusion: Captures both biological signatures of drugs and their topological relationships within the network.
• Performance: Demonstrated superior performance compared to previous state-of-the-art approaches and several classic machine learning methods.
• Output: Produces ranked candidate drugs for breast cancer treatment based on integrated network and signature information.

Scientific Applications:

• Drug repurposing for breast cancer: Predicts and ranks candidate existing drugs for potential therapeutic use in breast cancer.
• Literature-supported candidate identification: Identifies high-ranked drugs that have been reported in the literature as having therapeutic relevance to breast cancer.
• Support for therapy development: Provides prioritized candidate drugs to inform ongoing breast cancer therapy research and validation efforts.

Methodology:

Integration of drug network-based and drug signature-based analyses achieved through a GraphSAGE-based graph neural network that uses differentially expressed genes associated with breast cancer, drug-exposure gene expression data, and drug-drug link information to capture biological signatures and topological relationships.