Interpretaion of dendrogram height in WGCNA?

Question

[huynguyen96.dnu]

Hi there, I successfully constructed a gene coexpression network using the package WGCNA. Everything is okay. But right now, I want to get a deeper understanding of the output of this tool. Specifically, When I received a output something like:

As you can see, there are lot of branches with different heights. So what the difference between a longer branch's height [red rectangle] and a shorter branch's height [purple rectangle] ? And is it normal or not that dendrogram height can above 1? thank in advance!

Answer 1

[Kevin Blighe]

The heights relate to the distance metric that you used for clustering. As you are using WGCNA, the distance metric is likely One minus Pearson correlation (1 - r) due to the fact that the default parameter for this that is passed to the blockwiseModules() function is corType = "pearson".

So, the samples in the blue box are highly correlated to all of those other samples at the top of the dendrogram. Their tree height difference is roughly 1.0 - 0.9 = 0.1. Translated into Pearson, this is 1 - 0.1 = 0.9, i.e., highly correlated.

Those samples at the very bottom are weakly correlated to all of those in the tree structure at the top - 1 - (1.0 - 0.3) = 0.3

Kevin

[huynguyen96.dnu]

Many thanks for your helpful answer. I got additionally one question for u, sir. Is it normal or not that dendrogram height can above 1? I did follow a step-by-step tutorial. Something like that: https://imgur.com/vJKP0VA

[Kevin Blighe]

Yes, I believe, when you are using a signed WGCNA network, because Pearson correlation values can be negative, too. You should check the default values of your functions in order to understand the distance metric used. For blockwiseModules(), just type the following at the command prompt:

?blockwiseModules

[huynguyen96.dnu]

OMG, I forgot that Pearson corelation values can be negative. Thanks for your helps again!