WGCNA module-trait correlation heatmap has horizontal streaks -- how to interpret/fix?

Question

[janinubinu]

Hi,

I want to correlate my WGCNA modules with some other module eigengenes and used WGCNA's moduleTraitCor() function to do so. The correlations returned are shown in the heatmap below. WGCNA modules are on the vertical axis, my other modules are on the horizontal axis.

Can anyone give me any leads as to why the correlations look like this? I didn't expect some WGCNA modules to correlate with ALL blood modules. I am trying to figure out whether this is real or if there is something wrong with the way I set things up.

[andres.firrincieli]

What are the blood modules? How did you get these modules? what are the actual values of the blood modules used in moduleTraitCor()?

Answer 1

[LChart]

For one thing, the implication is that your "other modules" (horizontal axis) are all extremely highly correlated among themselves, such that a positive correlation between WGCNA.ME{i} and Blood.ME{k} strongly implies a positive correlation between WGCNA.ME{i} and Blood.ME{n} for every n. That is, the "banding" you're worried about suggests that the blood modules all have roughly the same expression pattern for all samples. What does the Blood x Blood heatmap look like? I'm willing to bet the median r is about 0.7 or 0.8.