How to calculate eigengenes on arbitrary expression data, not inside WGCNA

Question

[tospo]

I would like to compare the results from different algorithms for clustering data from an RNAseq experiment. One methodology we used is WGCNA, which represents each cluster/module by a module eigengene.

What I am now looking for is a way to calculate such eigengenes on arbitrary lists of genes and their expression profiles. I want to use that to reduce the clusters generated by other methods to a set of "representative genes" and compare that to the WGCNA output.

Unfortunately, I'm not clear about how to get from a list of genes and their expression profiles to an eigengene. The functions inside the WGCNA R package are deeply tied into the WGCNA analysis and I can't see how to use them on arbitrary data frames of gene expression data. Any hints would be highly welcome!

Answer 1

[Jean-Karim Heriche]

Eigengenes were first defined in this paper. Singular value decomposition (SVD) is what you're looking for. The eigengenes are the right singular vectors of the SVD of the expression matrix. If X is your data with genes as rows and samples as columns, the SVD of X is X=USV' and the eigengenes are defined as the vectors in V. In R,

eigengenes <- svd(X)$v

EDIT: Fixed link to paper.

[tospo]

thank you, I will try that!

[tospo]

BTW: the link to "this paper" doesn't work

[Jean-Karim Heriche]

Fixed. Sorry about that.

[m.sadman.sakib]

Thank you very much for this!!! I was looking for a similar solution for months and somehow just found this one line of code that solved my problem!