WGCNA heatmap modules-trait relationship

Question

[debitboro]

Hi everyone,

By using WGCNA for inference of coexpression networks I have generated the following values (heatmap) of correlation and significance of modules (10 modules), I mean the correlation between the modules and the trait of interest, the significance is measured by p_value

From the heatmap, we can say that the most three significant modules are: blue, black, and salmon modules. But, what about the green modules for example for the yellow module which is negatively correlated with the trait, is it also among the most significant modules like the red modules ?

Answer 1

[WouterDeCoster]

So your question essentially is whether anti-correlation is also of interest?

[debitboro]

yeah, this is my question

[WouterDeCoster]

I don't know your research hypothesis, but from a biological point of view a module which is anti-correlated with a trait is as interesting as a module which is correlated.

Say that your trait is weight, both modules which are overexpressed and those underexpressed with increasing weight are biologically interesting: which pathways are upregulated and which are repressed when you start devouring your 16th big mac of the day?

[zizigolu]

sorry,

in this image a trait such as Aortic.lession with red module has correlation 1 and p.value -0.019, but I can't figure out if correlation is 1 then why the color is not red (red color is correspond to positive correlation between module and trait).

http://g88i.imgup.net/Untitled4d01.png

[WouterDeCoster]

The p-value is 1, the correlation coefficient is -0.019.

As far as I know, p-value can't be negative. That doesn't make sense since it's a probability.

[zizigolu]

thank you. so helpful.