WGCNA what modules should contain ?

Question

[mchimich]

Dear users. I'm new the world of clustering of gene expression. I used WGCNA package and obtained different modules (using log2 transformed RPKM score of 3 different time point). However I'm still confused about the result obtained. When I draw the expression profile of genes contained in each module I obtained a kind of asymmetrical profile (whatever the threshold and parameters tested). for my understanding of gene expression clustering the two gene populations (see plot) that belong to a single WGCNA module must be classified in two different "modules". Does someone experience something like that before? Is it expected? Did I misunderstand the definition of module?

Thank you a lot for your help!

Answer 1

[Devon Ryan]

This becomes a question of whether or not one uses the absolute value of the covariation when making the modules. If you do take the absolute value, then you'll get graphs like this. I would personally argue that it makes biological sense to take the absolute value, since I think of modules as abstracted pathways and having both positively and negatively correlated changes in pathways makes complete sense.

[mchimich]

Thanks Devon for your answer. I did find the solution based on your explanation of the problem. this can be changed in WGCNA package by using signed network, here the explanation found:

In a signed correlation network, nodes with negative correlation are considered unconnected (their connection strength is zero or very close to zero). In contrast, in unsigned correlation networks, nodes with strong negative correlations have high connection strengths: the unsigned network adjacency is based on the absolute value of correlation, so positive and negative correlations are treated equally.

So to separate the two gene population I changed the network type to signed

adjacency = adjacency(datExpr, power = softPower,type = "signed")

Done!

Thanks again

[Devon Ryan]

+1 for finding the exact parameter. Best of luck!