WGCNA signed vs unsigned network

Question

[1769mkc]

How to choose between signed vs unsigned network ?

sft=pickSoftThreshold(datExpr,dataIsExpr = TRUE,powerVector = powers,corFnc = cor,corOptions = list(use = 'p'),networkType = "signed")

adj= adjacency(datExpr,type = "unsigned", power = softPower)

TOM=TOMsimilarityFromExpr(datExpr,networkType = "signed", TOMType = "unsigned", power = softPower)

These are the three lines where i have to defined signed or unsigned type i read through Peter Langfelder blog but still im not much clear about

Any suggestion would be highly appreciated

Answer 1

[Tao]

I think the "sign" represents the sign of weight on the edges. If you care about the sign, for example sign represents positive or negative regulation between two nodes, then you need to choose "signed". On the other hand, if the weight just represent the strength of relatedness between two nodes, you might don't care positive or negative, so you need to choose "unsigned". That really depends on your needs.

[1769mkc]

okay so i have cell lypes like stem cells and their downstream differentiated cells i m trying co-expression network .So what would you suggest me to do ?I have cell types like HSC,CMP,GMP,Monocyte, granulocytes etc..all together 20 samples..

Should I go for signed or unsigned ?

Any suggestion would be really helpful

[Tao]

For co-expression network, I would prefer unsigned weight. But you can try both, to find out which result looks more meaningful.