CWGCNA: an R package to perform causal inference from the WGCNA framework.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Weighted Gene Co-expression Network Analysis (WGCNA) is a powerful tool for identifying gene modules and their association with phenotypic traits.
• Existing WGCNA methods primarily focus on correlation, limiting the ability to infer causal relationships.
• There is a need for advanced analytical tools that can uncover causal links between gene expression patterns and biological outcomes.

Purpose of the Study:

• To develop an R package, CWGCNA (causal WGCNA), that extends traditional WGCNA by incorporating causal inference.
• To enable the determination of causal relationships between WGCNA modules, their features, and phenotypes.
• To introduce a novel network-based gene set function annotation method and machine learning capabilities for multi-omics data.

Main Methods:

• CWGCNA couples a mediation model with WGCNA to identify causal relationships.
• A network-based method is employed for gene set function annotation, considering module topology.
• The package includes machine learning functionalities for clustering and classification of multi-omics data.

Main Results:

• CWGCNA successfully identifies causal relationships between gene modules and phenotypes.
• The novel annotation method effectively captures the influence of module topology on gene set functions.
• Performance evaluation on three single and multi-omics datasets demonstrates superior results compared to existing methods.

Conclusions:

• CWGCNA provides a robust framework for causal inference in gene co-expression network analysis.
• The package offers enhanced capabilities for multi-omics data analysis, including causal discovery and machine learning.
• CWGCNA represents a significant advancement for researchers investigating gene-phenotype relationships and biological mechanisms.