-1
$\begingroup$

I tried to identify intramodular hub genes in WGCNA analysis. I used:

hub = chooseTopHubInEachModule(datExpr, moduleColors)

But when I checked the hub genes for the modules, I found that the hub genes are expressed only in less than 5 out of 300 samples. I did not filter the gene expression data for any low expressed genes in certain number of samples. How does this make biological relevance if the hub genes are not expressed in most of the samples?

Thank you!

$\endgroup$

1 Answer 1

1
$\begingroup$

It could be that your modules group together genes (over)expressed in only certain small groups of samples that may be biologically relevant. You would need to check whether the samples in which the hubs (and presumably the modules) are up share some sort of a characteristic.

In general though, there are good reasons why authors of WGCNA recommend filtering out non- and low-expressed genes - correlations tend to not be a good co-expression measure when counts are very low, and technical effects can be much more pronounced. It is also recommended to spend a bit of time on preprocessing to make sure the data set does not have strong global drivers that are technical or uninteresting.

$\endgroup$
1
  • $\begingroup$ thank you! I applied some filters on RNA-seq data input, gsg = goodSamplesGenes(GE.adjusted, minFraction = 1/2, minNSamples = 150, verbose = 3). I still have the genes which only have expression values for even one sample out of 300. I have a lot of zero's for no gene expression in the input file. Does WGCNA treat zero's as no or low gene expression? Should I replace zero's with NA so that it drop off the genes that have gene expression for less than say 100 samples? Thank you! $\endgroup$ Aug 31, 2021 at 21:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.