Linearity of network proximity measures: implications for set-based queries and significance testing

Area of Science:

• Bioinformatics
• Computational Biology
• Network Science

Background:

• Biomolecular interaction networks are crucial for understanding biological systems.
• Various network proximity measures exist for applications like disease gene prioritization and interaction prediction.
• A challenge lies in effectively scoring nodes based on their proximity to sets of 'seed' nodes.

Purpose of the Study:

• To propose a unifying framework for characterizing network proximity measures for set-based queries.
• To develop methods for processing set-based proximity queries using sparse local information.
• To provide an analytical framework for assessing the statistical significance of network proximity scores.

Main Methods:

• Characterized existing network proximity measures as linear aggregations.
• Developed methods for efficient set-based proximity query processing.
• Established an analytical framework using reference models to capture seed set characteristics for score distribution analysis.

Main Results:

• Proposed a unifying framework for network proximity measures.
• Enabled computation of exact statistical significance for network proximity scores.
• Facilitated accurate assessment of Monte Carlo simulation methods.

Conclusions:

• The proposed framework unifies diverse network proximity measures.
• It allows for accurate statistical validation of proximity-based findings in biological networks.
• Implementations and visualizations are available for practical application.