GLIDER

GLIDER transforms protein-protein interaction (PPI) and association networks into graph-based similarity networks to improve protein function prediction by preserving local graph structures while leveraging low-dimensional embeddings that group functionally similar proteins.

Key Features:

• Network transformation: Converts PPI and association networks into graph-based similarity networks that emphasize local neighborhood relationships.
• Embedding integration: Integrates embedding-based approaches with explicit graph structure preservation to retain topological information.
• Local and global properties: Captures implicit local and global network properties to reflect multiple scales of functional similarity.
• Low-dimensional embeddings: Produces low-dimensional embeddings that spatially group functionally similar proteins.
• Local neighborhood enrichment: Identifies local GLIDER neighborhoods that show significant functional enrichment.
• GO label prediction performance: Demonstrates superior performance in predicting Gene Ontology (GO) functional labels across multiple PPI networks.
• GLIDE lineage: Extends the GLIDE approach originally used for predicting missing links in protein-protein association networks.

Scientific Applications:

• Gene Ontology function prediction: Predicts GO functional labels across diverse PPI and association networks, including networks from the 2016 DREAM Disease Module Identification Challenge and STRING-derived networks.
• Disease-gene neighborhood analysis: Characterizes local neighborhoods of disease-related genes, exemplified by analyses of Parkinson's Disease GWAS genes to confirm pathway involvement and nominate novel candidate genes.
• Link prediction in association networks: Facilitates prediction of missing links in protein-protein association networks.

Methodology:

Transforms PPI/association networks into graph-based similarity networks and integrates embedding-based approaches with graph structure preservation to produce low-dimensional embeddings that capture implicit local and global network properties.