tools-only Co-expression Networks - Usage Guide
Overview
Build weighted gene co-expression networks to identify modules of co-regulated genes and relate them to phenotypes. WGCNA constructs a scale-free network from expression data, groups genes into modules based on co-expression patterns, and correlates module eigengenes with sample traits to find biologically relevant modules and hub genes.
Prerequisites
# R packages
install.packages('WGCNA')
install.packages('CEMiTool')
BiocManager::install('hdWGCNA')
# Python alternative
pip install PyWGCNA
Minimum 20 samples recommended for reliable module detection (absolute floor of 15).
Quick Start
Tell your AI agent what you want to do:
- "Build a co-expression network from my RNA-seq data"
- "Find gene modules correlated with treatment response"
- "Identify hub genes in the most significant module"
- "Run WGCNA on my bulk RNA-seq count matrix"
- "Find co-expression modules in my single-cell data"
Example Prompts
Module Detection
"I have normalized RNA-seq counts from 30 samples. Build a WGCNA network and find gene modules."
"Run CEMiTool on my expression data for an automated co-expression analysis."
Module-Trait Relationships
"Correlate my WGCNA modules with survival time and treatment group."
"Which modules are significantly associated with disease status?"
Hub Genes
"Find hub genes in the module most correlated with my phenotype."
"Export the turquoise module network for visualization in Cytoscape."
Single-Cell
"Run hdWGCNA on my Seurat object to find co-expression modules per cell type."
What the Agent Will Do
- Load and filter expression data for low-variance genes
- Select soft-thresholding power based on scale-free topology fit (R^2 > 0.85)
- Construct network and detect gene modules
- Calculate module eigengenes and correlate with sample traits
- Identify hub genes based on module membership and trait significance
- Visualize module-trait heatmap and export networks
Tips
- Sample size matters - WGCNA needs at least 15 samples, 20+ recommended for robust results
- Soft power selection - Choose the lowest power where scale-free R^2 exceeds 0.85; if it never reaches 0.85, aim for the plateau above 0.80
- Batch effects - Remove batch effects before network construction; batch-driven modules are artifacts
- Gene filtering - Use the top 5000-10000 most variable genes to reduce noise and speed computation
- CEMiTool for quick analysis - Automates soft-threshold selection and module detection; good for exploratory analysis
- hdWGCNA for single-cell - Creates metacells to address scRNA-seq sparsity before applying WGCNA
Related Skills
- scenic-regulons - TF regulon inference from scRNA-seq with pySCENIC
- differential-networks - Compare networks between conditions with DiffCorr
- differential-expression/deseq2-basics - DE analysis for gene prioritization
- differential-expression/batch-correction - Remove batch effects before network construction
- pathway-analysis/go-enrichment - Functional enrichment of gene modules