WGCNA, what does it mean if no hub genes are identified?
0
0
Entering edit mode
4 weeks ago
DNAngel ▴ 210

I ran WGCNA for my genes following the tutorials by Horvath and Langfelder (https://horvath.genetics.ucla.edu/html/CoexpressionNetwork/Rpackages/WGCNA/Tutorials/). I've obtained my modules and the GS and MM for my list of genes - however I think my situation is a bit odd. I've read in a few papers that to identify hub genes, one can use cut offs of absolute values of GS > 0.2, and MM > 0.8.

In my case, after implementing those cutoffs, nothing remains. All my genes had GS values JUST under 0.2, and MM just under 0.8. I am very confused and interested in this result - I don't know what this means. There were clearly modules with high significance (correlations were weak, about 0.3-0.7, but p values were quite significant <0.000001).

  1. Is the inability to find hub genes because my modules happen to have weak correlations with the traits of interest that I am testing? This is my only reasoning thus far as obviously no paper is going to talk about NOT finding any hub genes...

  2. Anyone aware of any other cutoffs papers have used? It seems that GS > 0.2, and MM > 0.8 is the standard but I think I remember reading a paper that only used GS > 0.2 although I can't seem to find it anymore.

WGCNA • 394 views
ADD COMMENT
1
Entering edit mode

There were clearly modules with high significance (correlations were weak, about 0.3-0.7, but p values were quite significant <0.000001)

can you plot the heatmap of the module with the highest module-trait relationship?

edit: also, can you post the chunk of code used to calculate the GS?

ADD REPLY
0
Entering edit mode

I can plot the heatmap but it doesn't really show the top module by trait very well. In fact I think the heatmaps are pretty unhelpful since almost always you have to run the GS and MM individually for each module to get the actual pvalues. Not sure why I need to post the code to calculate GS? It's just straight from the manual I posted in the link, my question is simply regarding their cut offs.

ADD REPLY
1
Entering edit mode

heatmaps are pretty unhelpful since almost always you have to run the GS and MM individually for each module to get the actual pvalues

I was asking for the heatmap because it sounds like the expression patterns in your modules are random: MM > 0.8 but GS < 0.2.

I was also asking for the GS calculation code because you mentioned modules with a significant 0.7 correlation in respect to the trait of interest. This is odd because within these modules I would expect to find hub genes (MM > 0.8) having a GS > 0.2 for that trait.

By the way, instead of using the GS, you could check if genes with a MM > 0.8 are also differentially expressed.

ADD REPLY
0
Entering edit mode

Oh yes good point. I may need play around with it a bit more - it is a few genes and I suspect I didn't have a large enough sample size as well which might be throwing it off. But your last point for differentially expressed seems to be a good way to go. Thanks!

ADD REPLY

Login before adding your answer.

Traffic: 1860 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6