Why does my WGCNA module appear as two groups in the heatmap?

Question

[LeeLee]

I want to perform WGCNA analysis on some proteomic data, after getting a series of modules I draw a heatmap for the specific modules, but I find that the genes in these modules are divided into two distinct groups in the heatmap. It is interesting that these two sets of genes seem to have a good negative correlation, I don't understand whether such a result is appropriate in WGCNA analysis?

Below is my WGCNA code:

> net = blockwiseModules(
>     datExpr,power = sft$powerEstimate,
>     maxBlockSize =nGenes ,
>     TOMType = "unsigned", minModuleSize = 40,
>     reassignThreshold = 0, mergeCutHeight = 0.2, 
>     numericLabels = TRUE, pamRespectsDendro = FALSE,
>     saveTOMs = F,deepSplit = 4,
>     verbose = 3)

[Sumit Paliwal]

Hi LeeLee, Did you do log transformation or any normalization of your raw data before performing WGCNA?

Answer 1

[andres.firrincieli]

Because by default WGCNA build an unsigned network and therefore negatively correlated genes are included in the same module.

source

[LeeLee]

Thanks for your reply, does this mean that if I want to separate the two negatively correlated group of genes, I just need to change TOMType = "unsigned" to "signed" ?

[andres.firrincieli]

First, you should recalculate power estimates with pickSoftThreshold by setting the argument networkType = "signed"

Finally,

net = blockwiseModules(
     datExpr,power = sft$powerEstimate,
     maxBlockSize =nGenes ,
     networkType = "signed" # adjacency matrix signed
     TOMType = "signed", # TOM signed
     minModuleSize = 40,
     reassignThreshold = 0, 
     mergeCutHeight = 0.2, 
     numericLabels = TRUE, 
     pamRespectsDendro = FALSE,
     saveTOMs = F,deepSplit = 4,
     verbose = 3)