Entering edit mode
4.9 years ago
anu014
▴
190
Hello all,
I wanted to know how beta is determined in order to calculate weighted network adjacency in WGCNA (Weighted correlation network analysis)?
Also, if I'm not wrong thresholding means below a particular value the entities, here nodes, are removed. But in WGCNA number of nodes remain the same even after Topology Overlap Matrix. Where is thresholding happening?
Please help me out in clearing my confusion.
Thanks
Please cite your references for
Oh yes, I forgot to add. Below are the references - 1. For 1st, https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-9-559 2. Yes, thresholding in terms of gene network analysis. 1st we deduce Pearson coefficient between gene pairs and then select a threshold of Pearson coefficient over which genes are considered to be connected or not (hard thresholding). But in case of soft thresholding we just convert the Pearson coefficient which has beta as exponent. I'm not understanding without removal of genes how the connection between genes are deduced
I do not know the 'intricacies' of WGCNA because I do not use it too much. Hopefully the answers that I have posted below can help you to understand the soft thresholding part?
Yes, in a simplistic network analysis, we can just remove edges that fall below a certain Pearson or Spearman correlation value (or whatever value is represented by the edge weights, which may be Euclidean distances, etc). This is what I do here: Network plot from expression data in R using igraph
In WGCNA all nodes are connected however, by raising the pearson coefficient to beta you change the connection strength so you can suppress low correlations (noise) while emphasizing the strong ones.
What is beta? The publication says it's calculated according to scale free Topology criteria. I don't understand this.
to a power (beta); function 5 in the manuscript.