Entering edit mode
I am performing a transcriptomic study of case-control samples. I need to identify genes that are key players in either case or control. Differentially expressed genes is the simplest approach for this. However, they do not take into account the interaction between the genes. Can WGCNA be used for the purpose? On performing WGCNA is it that genes responsible for either case or control are grouped into different modules? Post WGCNA, how can can I relate which module is for case condition and which for control?
To apply WGCNA it is recommended to have many samples. How many samples do you have for your case and for your control? WGCNA could be used for this, but the genes might not be on different modules as a gene might be involved in the transition from your control to your cases. I would recommend to perform WGCNA separately by case and controls and then asses if the genes of one module are on the same module on the other network. To relate modules to your phenotype WGCNA vignettes show a correlation between the egienvalue of the module and the phenotype.
Total 40. 20 for each case and control.
Of the ~16K genes I am analysing, ~10K genes are in a single module that is significantly negatively correlated to the sample type (case or control set as 0 or 1). And this raises all my doubts and difficulties.
So as per your suggetion, I must follow Tutorial II from the WGCNA website instead of Tutorial I which I am currently following.
Does it mean module-trait relationship analysis?
Having a big module is expected if the network follows the power law (which WGCNA expects the data to follow).
Yes, the module-trait relationship is what I meant. But you seem to have run it already but perhaps with all the samples together. See if following the second tutorial results in something different. I expect it will, but if not how many more traits do you have? Are there any confounders ?
I actually donot have any trait data. For sample type as trait data I used 0 and 1 for case and control.
There were confounders which I have already removed using
limma::removeBatchEffect()and then transformed my data usingvst()