A strategy to incorporate prior knowledge into correlation network cutoff selection.

Area of Science:

• Systems biology
• Bioinformatics
• Network science

Background:

• Correlation networks are common for identifying biological interactions from omics data.
• Statistical significance of correlations is typically used for network edge selection.
• This statistical approach may not accurately capture underlying biological mechanisms.

Purpose of the Study:

• To develop and evaluate an alternative network reconstruction approach.
• To improve the biological relevance of inferred networks by integrating prior knowledge.
• To assess the generalizability of the method across different omics datasets.

Main Methods:

• A cutoff selection algorithm was developed to maximize network overlap with prior biological knowledge.
• The approach was evaluated on IgG glycomics, metabolomics, and transcriptomics data.
• Performance was compared against networks derived from statistical significance.

Main Results:

• The proposed method successfully reconstructed biologically relevant networks.
• The algorithm demonstrated robustness even with incomplete or inaccurate prior knowledge.
• Optimized networks from transcriptomics data retrieved more biologically relevant interactions than statistical networks.

Conclusions:

• The novel network reconstruction method enhances the discovery of biological interactions.
• Integrating prior knowledge into network inference is a powerful strategy for omics data analysis.
• This approach offers a superior alternative to purely statistical methods for biological network analysis.