gene_module_score: Gene module score

In vittoriofortino84/COPS: Clustering algorithms for Omics-driven Patient Stratification

Gene module score

Description

Metric based on gene module eigen gene correlation

Usage

gene_module_score(
  x,
  module_eigs,
  module_cor_threshold = 0.3,
  module_nan.substitute = 0,
  ...
)

Arguments

x	A data.frame with columns "id" and "cluster".
module_eigs	Gene module eigen-genes for each sample (samples x modules).
module_cor_threshold	Threshold for counting correlations.
module_nan.substitute	Substituted value when dividing by zero when there are no correlated clusters for a module.
...	Extra arguments are ignored.

Examples

library(COPS)

# Generate module eigen genes with WGCNA
gene_correlation <- cor(t(ad_ge_micro_zscore), method =  "spearman")
adj <- abs(gene_correlation)^2
TOM <- unsigned_TOM_matrix(adj)
geneTree <- flashClust::flashClust(as.dist(1 - TOM), method="average")
dynamicMods <- dynamicTreeCut::cutreeDynamic(
  dendro = geneTree,  
  method="tree", 
  minClusterSize = 20
)

MEs <- gene_module_eigengenes(
  ad_ge_micro_zscore[dynamicMods != 0,], 
  dynamicMods[dynamicMods != 0]
)

# Compute the module score for a given clustering result
clust <- cutree(
  hclust(
    as.dist(1-cor(ad_ge_micro_zscore, method = "spearman")), 
    method = "average"
  ), 
  k = 3
)
clust <- data.frame(id = names(clust), cluster = clust)

score <- gene_module_score(clust, MEs)

# Within full pipeline
res <- COPS(ad_ge_micro_zscore, 
association_data = ad_studies, 
parallel = 1, nruns = 2, nfolds = 5, 
dimred_methods = c("pca", "umap"), 
cluster_methods = c("hierarchical", "kmeans"), 
distance_metric = "euclidean", 
n_clusters = 2:4, 
module_eigs = MEs)

scores <- scoring(res, wsum = Silhouette - Module_score, summarise = TRUE)

vittoriofortino84/COPS documentation built on Jan. 28, 2025, 3:16 p.m.