Question about donor pseudobulks and donor-level ME values

[svashishtphd]

Hello! I have a general question about the MEs that I obtain from pseudobulked hdWGCNA. I followed the vignette and GetMEs yields cell-level ME values rather than donor level values. My psuedobulks are donor_cluster and that is the level for which I want to see MEs and use this for downstream Module-Trait correlations. Ideally, I want one measure per module for each donor to carry into analysis. Should my hdWGCNA input be at an aggregated level prior to using the tool, because currently I am using the package to achieve aggregation. I attempted to calculate donor-level MEs using the function below but I am curious if there is an easier/better way to obtain a value for each module by donor rather than cell-level.

get_pseudobulk_MEs <- function(seu_obj, prefix) {

• datExpr <- GetDatExpr(seu_obj)
• datExpr <- as.data.frame(datExpr)
• mods <- GetModules(seu_obj)
• mods <- mods[mods$module != "grey", , drop = FALSE]
• gene_col <- if ("gene_name" %in% colnames(mods)) "gene_name" else "gene"
• gene2mod <- setNames(as.character(mods$module), as.character(mods[[gene_col]]))
• common_genes <- intersect(colnames(datExpr), names(gene2mod))
• datExpr_use <- datExpr[, common_genes, drop = FALSE]
• mod_colors <- gene2mod[common_genes]
• me_list <- WGCNA::moduleEigengenes(

• expr = as.matrix(datExpr_use),
  

• colors = mod_colors
  

• )
• me_df <- as.data.frame(me_list$eigengenes)
• rownames(me_df) <- rownames(datExpr_use)
• colnames(me_df) <- sub("^ME", "", colnames(me_df))
• colnames(me_df) <- paste0(prefix, "_", colnames(me_df))
• me_df$pb_id <- rownames(me_df)
• me_df$Participant_ID <- sub("\..*$", "", me_df$pb_id)
• me_df

[smorabit]

In general, hdWGCNA provides cell-level outputs even when we use a metacell or pseudobulk representation to calculate the network. I did not try to run your code, but it does seem reasonable. Maybe in the future I could add an advanced tutorial for doing all of the downstream analysis with a pseudobulk dataset, since it should be possible (albeit with a lot of custom code).

hdWGCNA does already calculate these values and you can access them like this:

net <- GetNetworkData(seurat_obj)
MEs <- net$multiMEs[[1]][[1]]

Note that the rownames of this table will correspond to the original module names, so if you use the hdWGCNA functions to re-name the modules, they will not match.