Disparity-filtered differential correlation network analysis: a case study on CRC metabolomics

Area of Science:

• Systems Biology
• Bioinformatics
• Metabolomics

Background:

• Differential network analysis reveals changes in interactions across conditions, offering insights into dysregulated pathways and biomarkers.
• Current methods for detecting differential interactions are diverse and often rely on flawed assumptions.
• Establishing direct links between observed interactions and biochemical mechanisms remains challenging.

Purpose of the Study:

• To introduce a novel method for differential network analysis that overcomes limitations of existing approaches.
• To develop a classification model based on inferred network interactions for disease identification.
• To provide a robust framework for analyzing metabolomics data and interpreting biological pathways.

Main Methods:

• A new differential network analysis method employing a disparity filter for network reduction.
• Statistical significance testing against a null model without favoring specific resolution scales or making Gaussian assumptions.
• Development of a classification model utilizing the inferred network interactions.

Main Results:

• The method successfully identified significant differential interaction patterns in a colorectal cancer (CRC) metabolomics dataset.
• Key metabolites identified as hubs align with existing literature findings.
• The classification model accurately differentiated between colorectal cancer, polyp, and healthy subjects.

Conclusions:

• The proposed method offers a simple and effective framework for identifying differential interaction patterns.
• This approach enhances the biological interpretation of metabolomics data.
• The method provides a valuable tool for biomarker discovery and disease classification.