GLIDE

GLIDE integrates a generalization of the diffusion state distance with local network measures to predict missing links in protein-protein interaction (PPI) and other biological networks.

Key Features:

• Embedding-Based Methodology: Utilizes a generalization of the diffusion state distance to capture complex network structure.
• Integration of Global and Local Structures: Combines global network embeddings with customized local measures tailored to specific biological network types.
• Network Type-Specific Customization: Adapts to differing dominant structures across networks, identifying simple local measures as predominant in highly connected core regions between hub genes while relying on global embeddings elsewhere.
• Rigorous Validation: Validated on human networks from the 2016 DREAM disease module identification challenge and on a classical yeast PPI network using cross-validation.
• Practical Applications: Generates novel link predictions, for example proposing associations between genes implicated in Crohn's disease and genes not previously linked, supported by additional network data and literature.

Scientific Applications:

• Disease Module Identification: Predicts missing links to aid identification of disease modules and elucidate molecular mechanisms underlying diseases.
• Protein-Protein Interaction Networks: Predicts unobserved interactions in PPI networks to reveal new functional associations and pathways.

Methodology:

Computes node embeddings using a generalization of the diffusion state distance and integrates these global embeddings with local connectivity measures to score and predict missing links.