Integration strategy is a key step in network-based analysis and dramatically affects network topological properties...

Area of Science:

• Systems Biology
• Bioinformatics
• Network Science

Background:

• Molecular interaction networks are vital for understanding biological processes and disease mechanisms.
• Current network construction relies on integrating data from various sources due to data incompleteness and false positives.
• Existing integration strategies lack rigorous evaluation regarding their impact on network topology and applications.

Purpose of the Study:

• To systematically evaluate 11 common methods for integrating protein-protein interaction data.
• To compare the topological properties of networks generated by different integration strategies.
• To assess the impact of these strategies on key biological applications.

Main Methods:

• Analysis of 11 frequently used integration methods.
• Categorization of methods into empirical and machine learning strategies.
• Evaluation of resulting network topological properties.
• Assessment of network performance in disease gene prediction, drug target identification, and molecular complex detection.

Main Results:

• Significant differences in topological properties were observed among networks generated by various integration strategies.
• The choice of integration strategy profoundly affected the outcomes of disease gene prediction, drug target identification, and molecular complex detection.
• Empirical and machine learning methods yielded distinct network structures and application performances.

Conclusions:

• The integration strategy is a critical determinant of molecular network quality and utility.
• Network topology significantly influences the success of network-based biological applications.
• This study provides a foundation for selecting optimal integration strategies in future network biology research.