I was wondering if anyone could tell me how it would be possible to have genes that have negative correlations (as determined by kME values) with a module in a signed WGCNA analysis. My understanding is that, by definition, all the genes in a module should have a positive correlation with the module it is in if a signed analysis was done. If it helps, the code I used to generate the modules was:
net = blockwiseConsensusModules(maxBlockSize = 15000, multiExpr, power = 16, TOMtype = "signed", minModuleSize = 30, deepSplit = 2, pamRespectsDendro = FALSE, mergeCutHeight = 0.25, numericLabels = TRUE, minKMEtoStay = 0, saveTOMs = TRUE, verbose = 5)
I'm wondering what could be going on here. This didn't occur when I ran WGCNA on the sets of data from which the consensus was built individually. It's also worth noting that the kME values of the genes are always negative in both sets of data I used in the consensus, so it isn't as though it's positive in one and negative in the other.
Any input here would be greatly appreciated, as I don't feel comfortable moving on in my analysis until I understand this.