0
$\begingroup$

We are working with a lot of samples, more than 50,000 people! We are using WGCNA to get modules and it seems like there are less number of module genes the more people we use. We are following the tutorial guidelines. We set the min modulesize = 30 and deepSplit = 4 (max sensitivity) However, we are getting very small modules with numbers almost 30 for all. We want to know if using this many samples creates an artifact and the variation between the samples may reduce the correlation seen between the genes? If that is the case, do you suggest any way we can fix that? Also do you think z-scaling expression data improves the module detection?

Thank you so much.

$\endgroup$
2
  • $\begingroup$ Weighted correlation network analysis would require a clear understanding of the a priori weightings in my opinion. The background is here $\endgroup$
    – M__
    Nov 13, 2020 at 20:43
  • $\begingroup$ Where can access this background on priori weightings you suggested? Thanks! $\endgroup$
    – Habesha
    Nov 18, 2020 at 18:22

1 Answer 1

1
$\begingroup$

The biggest difference a large number of samples makes is that you can usually decrease the soft thresholding power. If you use power 6 (or 12 for a signed network), try decreasing it to 3 or even 2 (6 or 4 for signed network). Apart from that, it could also be that the co-expression modules in your data are simply smaller than what people usually find, nothing apriori wrong with that.

$\endgroup$
2
  • $\begingroup$ Okay, thank you. I am already using 2 as a threshold. I am afraid there may be confounding factors between the samples that are making it difficult for detection of modules. Are there confounding factors that I should be looking at removing that are known to affect module detection? $\endgroup$
    – Habesha
    Nov 18, 2020 at 18:22
  • $\begingroup$ I cannot make a definite statement, it depends on the minutiae of the data. You could certainly try some noise filtering approaches although I have yet to see a bulk (not single cell) sequencing data set where these help. $\endgroup$ Nov 30, 2020 at 7:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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