km2gcn

km2gcn refines gene co-expression network modules by applying k-means clustering to Weighted Gene Co-expression Network Analysis (WGCNA) outputs to improve module accuracy and biological relevance.

Key Features:

• Integration with WGCNA: Builds on WGCNA to construct gene co-expression networks and initial modules.
• k-means clustering enhancement: Applies k-means clustering as a supplementary step to reassign genes and refine module boundaries.
• Reduced misplacement of genes: Minimizes incorrect gene assignments to modules, aiming for few or zero misplaced genes.
• Increased replicability: Increases replicability of clusters across tissues with an average improvement factor of 3.1.
• Enhanced biological enrichment: Improves Gene Ontology enrichment (observed in 48 of 52 GCNs) and increases cell type enrichment signals in brain networks (improved in 21 of 23 cases).
• Accuracy in simulated data: Shows superior partition accuracy on simulated networks derived from GTEx data as measured by similarity indices.
• Validation datasets: Evaluated on UKBEC data from 10 human brain tissues and on GTEx data covering 42 human tissues (including 13 brain tissues).

Scientific Applications:

• Tissue-specific gene expression analysis: Enables more accurate module identification in studies of tissue-specific gene expression, particularly in complex tissues such as the human brain.
• Downstream functional interpretation: Supports downstream analyses that require biologically meaningful modules, including Gene Ontology and cell type enrichment studies.

Methodology:

Generate gene co-expression networks using WGCNA, apply k-means clustering to refine modules, and evaluate module properties using similarity indices, Gene Ontology enrichment, and cell type enrichment on UKBEC, GTEx, and simulated GTEx-derived networks.