A pseudo-value regression approach for differential network analysis of co-expression data.

Area of Science:

• Bioinformatics
• Systems Biology
• Network Analysis

Background:

• Differential network (DN) analysis reveals changes in gene association measures across experimental conditions.
• Existing methods lack the ability to adjust for clinical covariates.

Purpose of the Study:

• Introduce PRANA (pseudo-value regression approach for network analysis), a novel DN analysis method.
• Enable DN analysis to adjust for clinical covariates.
• Compare PRANA's performance against existing methods.

Main Methods:

• Utilizes mutual information criteria and pseudo-value calculations.
• Employs a robust regression model.
• Incorporates adjustment for clinical covariates.

Main Results:

• PRANA demonstrates superior performance in precision, recall, and F1 score compared to dnapath and DINGO.
• PRANA effectively adjusts for covariates like patient age, which other methods do not.
• PRANA successfully identified differentially connected genes associated with chronic obstructive pulmonary disease in a real-world dataset.

Conclusions:

• This study presents the first regression modeling approach for DN analysis that includes clinical covariates.
• Adjusting for covariates significantly enhances the accuracy of differential network analysis.
• PRANA offers a more robust and accurate method for identifying biologically relevant gene network changes.