how to treat the replicates while performing WGCNA

Question

[mavy]

Hello All,

I have RNAseq data, sensitive and tolerant to a drug , which is treated with one drug so I have four different classes of data 1)sensitive+treated,2)sensitive+untreated , 3)tolerant+untreated and 4) tolerant +treated each having three replicates

I want to perform WGCNA to find a module or two that is unique in either sensitive or resistant i.e the genes that are differentially expressed in sensitive and not in resistant or vice versa. I have three replicates of each sample, so my question is how to handle the replicates because if I am considering as R1,R2,R3 they are having different module eigen values so how can I merge them to have one representative module

Should I merge them before WGCNA or after and how? I am new to bioinformatics this may be basic but would really appreciate any help

[andres.firrincieli]

You do not merge the MEs. It is quite normal that some modules have noisy eigengenes (MEs); WGCNA is an unsupervised analysis overall.

By looking at the MEs, the Green module appears to be negatively correlated with the Sen.UT condition, meaning that the genes in this modules have lower expression in this group of samples compared to all the other conditions. The Blue module instead seems to include genes more expressed in the Tollerant (X1) genotype while being expressed at lower level in the Sensitive one.

[mavy]

Thankyou so much Andres for your reply. I think I got your point, just have a couple of quick questions to make it more clear .

1) I do not want to merge the MEs rather I asked about replicates, as some replicates of same sample are showing different behavior for the same Module

2) From your answer am concluding that rather than looking for higher ME values I should focus on a consistent value through out the sample replicates like as you mentioned blue module has all positive value for the tolerant one and -ve for the sensitive one ??

[andres.firrincieli]

1) I do not want to merge the MEs rather I asked about replicates, as some replicates of same sample are showing different behavior for the same Module

Don't merge the replicates. It is what it is; that noise is perfectly fine.

2) From your answer am concluding that rather than looking for higher ME values I should focus on a consistent value through out the sample replicates like as you mentioned blue module has all positive value for the tolerant one and -ve for the sensitive one ??

Exactly!

[mavy]

Thank you so much for your response . Really appreciate it. I will try the way you mentioned and will get back if I have any query