What is the need for signed or signed hybrid (as in WGCNA) co-expression networks?

Question

[Arindam Ghosh]

Since co-expression networks are build to understand the patterns of interaction/co-expression between genes, the unsigned version seems to be the best option as it disregards the sign of correlation value and can show interactions even between oppositely regulated genes. However, in case of signed-hybrid network, the negative co-relations are treated as zero. This limits the regulation between oppositely regulated genes in the network. So why is this desirable?

Answer 1

[andres.firrincieli]

Hi

Signed or unsigned which network type is preferable?

Signed network from signed topological overlap

Signed and signed hybrid: what’s the difference?

[Arindam Ghosh]

Well, I have read the contents of this blog. As it explains or rather recommends signed network due to two reasons:

First, more often than not, direction does matter: it is important to know where node profiles go up and where they go down, and mixing negatively correlated nodes together necessarily mixes the two directions together. Second, negatively correlated nodes often belong to different categories. For example, in gene expression data, negatively correlated genes tend to come from biologically very different categories.

But there are also sort of disclaimers:

It is true that some pathways or processes involve pairs of genes that are negatively correlated; if there are enough negatively correlated genes, they will form a module on their own and the two modules can then be analyzed together.

and

By and large does not mean always, and there may be applications in which an unsigned network is preferable. In principle there’s also nothing wrong with carrying out both types of analysis, but working with two related yet distinct analyses of the same data may quickly get confusing and tiring.

"there may be applications in which an unsigned network is preferable" in which situations will they be preferable.

[andres.firrincieli]

The first disclaimer pose a 'biological' question. If you think is more informative keeping strong negatively correlated genes in the same module, it's entirely up to you. However, with 'unsigned' networks it could be difficult to understand which hub genes in my module negatively/and positively correlate with your categorical/continuous variables. This is why I always prefer 'signed' network; I can easily find which hubs positively/negatively correlate with my experimental variables.

In regard to the second disclaimer I can not figure out in which situation an 'unsigned' network is preferred over a 'signed' or 'signed-hybrid' network. It could also be a 'biological' question.