WGCNA: Weighted Gene Co-expression Network Analysis in R

The 10 Modular Training Chapters

Chapter 1: Conceptual Framework: Differential Expression vs. Co-expression Networks in Systems Biology

Pairwise differential expression versus systems-level co-expression topology; biological foundations of scale-free networks; module definitions, module eigengene summarization, and diagnosing the grey module without dogmatic percentage thresholds.

Chapter 2: Bioconductor Stack Setup, Environment Isolation & Input Data Contracts

Installing WGCNA, DESeq2, AnnotationDbi, and plant annotation packages (org.At.tair.db); enforcing strict data contracts between raw integer count matrices and multi-factor experimental metadata; defensive assertions in R.

Chapter 3: Count Filtering, Variance-Stabilizing Transformation (VST) & Sample Outlier QC

Biological count filtering across sample replicates, applying DESeq2 variance-stabilizing transformation (vst) with blind=FALSE, transposing to sample-by-gene format, and running goodSamplesGenes with hierarchical outlier clustering.

Chapter 4: Soft-Threshold Selection & Scale-Free Topology Fitting

Mathematical derivation of soft vs hard thresholding; evaluating signed network adjacencies; executing pickSoftThreshold; plotting scale-free fit vs mean connectivity; diagnostic flowchart for handling NA power estimates with scientific justification.

Chapter 5: Network Construction, Topological Overlap Matrix (TOM) & Dynamic Module Detection

Constructing co-expression networks with blockwiseModules; calculating signed Topological Overlap Matrices (TOM); dynamic tree cutting; multi-block handling; converting numeric labels to colors; and merging similar modules at mergeCutHeight = 0.25.

Chapter 6: Module–Trait Associations, Matrix-Wide BH FDR Correction & Linear Modeling

Calculating Module Eigengenes (MEs); computing Pearson correlations against continuous and binary physiological stress traits; applying Benjamini-Hochberg FDR correction across the entire test matrix; and fitting linear models (lm(ME ~ treatment * timepoint)).

Chapter 7: Intramodular Hub Gene Prioritization & Module Membership (kME) Analysis

Quantifying module membership (kME) as correlation between gene expression and module eigengenes; calculating Student t-test p-values and BH-adjusted q-values; ranking central hubs; and annotating with Arabidopsis TAIR locus descriptions.

Chapter 8: Functional Enrichment with clusterProfiler & Statistical Background Definition

Performing Gene Ontology (GO) Biological Process and KEGG over-representation analysis; avoiding false positive inflation by properly defining the gene universe as non-grey, mapped network genes; semantic reduction with simplify; generating publication dotplots.

Chapter 9: Memory-Conscious Cytoscape Export & Interactive Network Visualization

Memory calculations for dense Topological Overlap Matrices (N^2 * 8 bytes); safe intramodular subnetwork extraction; subsetting to top hub genes (MAX_NODES = 300); exporting weighted edge and node files with exportNetworkToCytoscape; Cytoscape styling guide.

Chapter 10: Network Validation, Module Preservation, Troubleshooting & Experimental Follow-up

Validating co-expression modules across independent datasets with modulePreservation; troubleshooting common WGCNA pitfalls (batch effects, soft power anomalies, large grey modules); designing follow-up experimental perturbations (CRISPR/T-DNA mutants); Quarto workflow.

🎯 Target Audience & Prerequisites

• Plant & Agricultural Biologists: Researchers dissecting drought, heat, salinity, and biotic stress transcriptomes in crop and model species.
• Bioinformaticians & Data Scientists: Analysts transitioning from pairwise DESeq2 lists to higher-order modular co-regulation architectures.
• Systems Geneticists: Scientists mapping expression quantitative trait loci (eQTLs) and correlating module eigengenes with phenotypic trait measurements.
• Prerequisites: Basic familiarity with R/RStudio syntax and high-throughput RNA-seq count tables.

🏆 Key Practical Competencies

• Data Preprocessing & Homoscedasticity: Variance-stabilizing transformation with blind = FALSE and sample outlier pruning with goodSamplesGenes.
• Scale-Free Network Topology: Evaluating signed adjacency functions, plotting scale-free fit ($R^2 \ge 0.85$), and solving NA power estimates.
• Rigorous Trait Association: Pairwise Pearson correlation with observation tracking and matrix-wide Benjamini-Hochberg FDR correction.
• Background-Controlled Functional Mining: Preventing false positive inflation by restricting Gene Ontology universe strictly to expressed non-grey network genes.