Question: WGCNA network with condition and normal samples
gravatar for saptarathideb
17 months ago by
saptarathideb0 wrote:

How does WGCNA generate co expression network taking both normal and conditional samples as input in one go? I cant understand how correlations are generated in such case?

rna-seq network wgcna • 1.1k views
ADD COMMENTlink modified 17 months ago by Jake Warner780 • written 17 months ago by saptarathideb0
gravatar for Jake Warner
17 months ago by
Jake Warner780
Jake Warner780 wrote:


WGCNA is unsupervised in the network construction and phenotypic data is used for post hoc correlation tests. So in basic terms all the expression data is fed in at once and if a group of genes responds specifically to one treatment they should be detected as a module and correlate with the sample group. Meanwhile assuming you have enough samples other modules of genes that co-express independently of your treatment will also be identified.

ADD COMMENTlink written 17 months ago by Jake Warner780

Thanks a lot for the answer, Just one more thing if you can kindly answer:- If WGCNA calculates the correlation matrix without differentiating among sample types(normal Vs condition), I feel the correlation will not be good as it should be since many genes which are differentially expressed will have high impact on correlation. Please correct me...I am getting a blurred understanding but not getting it concretely.. Thank you

ADD REPLYlink written 17 months ago by saptarathideb0

You are correct, the genes which are regulated by the treatment will move as a group and show up as a module. But even if most of the genes move they wont move in the same way and you will detect multiple modules (I'm oversimplifying a bit here). A big factor is the number of samples overall... If you only have three controls and three treatments you will find few, large modules that correlate very strongly with treatment. But if you have many samples (the developers recommend >16) you will detect modules that are less dependent on phenotype/treatment.

ADD REPLYlink written 17 months ago by Jake Warner780

Thank you sir, So, basically module detection is not reliable when sample size is very low (ie. less than 15 sample per condition), right?

ADD REPLYlink written 17 months ago by saptarathideb0

Is it possible to identify which module corresponds to CASE or CONTROL? For module-trait relationship, I used 1 and 2 to denote CASE and CONTROL. I obtained the modules which are positively or negatively related to the trait but unsure which module relates to which group.

ADD REPLYlink modified 28 days ago • written 28 days ago by Arindam Ghosh200
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1532 users visited in the last hour