Question: WGCNA - Signed vs signed hybrid networks
gravatar for bram.verstockt
16 months ago by
KUL, Belgium
bram.verstockt0 wrote:

Dear all

As I just started using WGCNA, it's not entirely clear for me what's the difference between "signed" and "signed hybrid" networks. Can anyone clarify, please? 

Many thanks in advance!



ADD COMMENTlink modified 16 months ago by Peter Langfelder1.3k • written 16 months ago by bram.verstockt0
gravatar for Peter Langfelder
16 months ago by
United States
Peter Langfelder1.3k wrote:

In practical terms, the difference is negligible.

In a signed network, the similarity is defined as (1+cor)/2; in a signed hybrid the similarity equals the correlation if the correlation is positive and zero otherwise (it is called hybrid because it is a hybrid of weighted and unweighted networks). The similarity is raised to a power which is usually twice as big in signed networks than it is in hybrid networks. For correlations near 1, ( (1+cor)/2 )^(2*beta) approximately equals cor^beta, and for low correlations both adjacencies are near zero.


ADD COMMENTlink written 16 months ago by Peter Langfelder1.3k

Dear Dr. Langfelder

Many thanks for your quick response! As the difference between signed and signed hybrid is neglibigle, is there a difference in their applications? For instance, do you recommend one of both when analyzing data in relation to a binary trait? Many thanks for your answer.

ADD REPLYlink written 16 months ago by bram.verstockt0

I usually use signed hybrid since I like to have adjacencies for negative correlations exactly zero and I find the formula cor^beta "cleaner", but as I said, these are more "cosmetic" rather than fundamental reasons.

ADD REPLYlink written 16 months ago by Peter Langfelder1.3k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 224 users visited in the last hour