Should reassign_modules be set to TRUE or FALSE in ModuleConnectivity() for pseudobulk hdWGCNA?

[pengjiao]

Hi, thank you for developing hdWGCNA.
I am using hdWGCNA for pseudobulk analysis, and I noticed that many genes are reassigned to different modules after running ModuleConnectivity(). Would you recommend using reassign_modules = FALSE or reassign_modules = TRUE in a pseudobulk hdWGCNA workflow?
Thank you so much.

Here is the relevant code:

obj <- readRDS("Vascu.rds")
obj <- subset(obj, subset = cogdx %in% c(1,4)) #select samples

obj <- SetupForWGCNA(
  obj,
  gene_select = "fraction",
  fraction = 0.05,
  wgcna_name = paste0(str_sub(ct,1,3), "_pseudobulk")
)

# get the counts matrix and the meta-data
X <- GetAssayData(obj, layer = "counts")
meta <- obj@meta.data

# create a pseudo-bulk SummarizedExperiment object
se <- AggregatePseudobulk(
  X, meta,
  replicate_col = "individualID",
  group_col = "cell_type",
  assay_name = "counts"
)

# normalize the pseudobulk SummarizedExperiment
se <- NormalizeCounts(
  se,
  method = "VST",
  assay_name = "counts"
)

# Set the pseudobulk matrix using the SummarizedExperiment object
obj <- SetDatExpr(
  obj,
  mat = se,
  layer = "VST"
)

# select the soft power threshold
obj <- TestSoftPowers(obj) #soft_power = 9
p <- PlotSoftPowers(obj)
cowplot::plot_grid(plotlist = p)

# construct the co-expression network and identify gene modules
obj <- ConstructNetwork(obj, tom_name = paste0(str_sub(ct,1,3), "_test1_pseudobulk"),)

PlotDendrogram(obj, main="after ConstructNetwork")

obj1 <- ModuleEigengenes(obj)
PlotDendrogram(obj1, main="after ModuleEigengenes")

obj2 <- ModuleConnectivity(obj1)
PlotDendrogram(obj2, main="after ModuleConnectivity")

[smorabit]

Hi, I am not sure if there is a generalizable answer to your question across all pseudobulk workflows, but it seems like in your case based on the dendrogram plots that you provide that you should consider setting reassign_modules = FALSE.

To clarify what this option actually does, the network and modules are calculated based on the pseudobulk representation, but to connect the analysis back to the original single-cell dataset, hdWGCNA calculates module eigengenes (MEs) and module connectivity (kME) at the cell-level. There can sometimes be discrepancies wherein a gene that is assigned to a particular module actually is anticorrelated with that module at the single-cell level, and in these cases, reassign_modules = TRUE calculates the module for which that gene has the highest correlation at the cell-level, and reassigns it to that module.