Gene expression data analysis using Hellinger correlation in weighted gene co-expression networks (WGCNA).

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Weighted Gene Co-expression Network Analysis (WGCNA) typically uses linear correlations to identify gene modules.
• Linear relationships may not fully capture complex, functional dependencies between genes.
• Existing methods may miss non-linear biological interactions.

Purpose of the Study:

• To compare various correlation methods for capturing gene dependence in WGCNA.
• To evaluate the performance of Hellinger correlation as a sensitive measure in WGCNA.
• To identify novel gene network relationships, particularly non-linear ones, in human brain tissues and disease states.

Main Methods:

• Compared six different correlation methods for gene pairwise dependence.
• Applied Hellinger correlation within WGCNA framework using RNA-seq and microarray data.
• Constructed gene co-expression networks from human frontal cortex (Alzheimer's disease), temporal cortex, prefrontal cortex (single-cell), and GTEx heart data.

Main Results:

• Hellinger correlation yielded similar results to linear methods but revealed additional functional relationships.
• Identified a novel link between inflammation and mitochondria function.
• Validated network constructions across multiple datasets, demonstrating robustness.

Conclusions:

• Hellinger correlation robustly detects non-linear, biologically meaningful gene relationships.
• This method offers a complementary and more flexible approach to WGCNA.
• Application of Hellinger correlation to WGCNA uncovers novel network interactions in gene expression analysis.