Alternatives to WGCNA

Question

[Poorvi]

I am currently working with a microarray dataset. I have used WGCNA on it to produce around 12 biologically meaningful modules. However, I would like to test the modules accuracy and significance by trying out an alternative clustering method on my data. I have tried to use Leigen, Louvain, Spectral clustering etc. But they all don't seem to form similar modules as those produced by WGCNA.

Can someone please help me out or point me to a tutorial which can produce similar and biologically significant modules like in WGCNA?

Answer 1

[LChart]

GLASSO, ARACNe-AP, and Megena are alternative co-expression frameworks that have been applied. However standard clustering methods should give you similar clusters to WGCNA when applied correctly. WGCNA is fundamentally correlation based (correlations are transformed into a topological overlap similarity metric) - so you need to set your distance to the cosine metric to ensure similar raw distances/similarities are being used.