Hi,

I am doing WGCNA on methylation data (EPIC) from human blood samples and I'm stuck when choosing the soft threshold. I already made WGCNA on RNA-seq data before and the step of choosing the soft threshold was easy, with a high SFT.R.sq and a low mean connectivity (0.9 and 48). I selected the 240 000 most variable beta-values, but my methylation dataset doesn't seem to follow a scale free topology (the SFT.R.square is <0.7 and the mean connectivity is equal to several thousands). I saw here and here that I should take a beta of 12 (I have 45 samples). However, before running blockwiseModules(), I would like to know if someone else met this type of difficulty and how do you managed it. Were your results relevant with your biological question ?

Thanks a lot in advance,

Mean connectivity in the several thousands is a bit too high; I would investigate whether the samples cluster in strong clusters and if so, would probably adjust for the cluster membership. You may also want to adjust for age and sex of the donors, if you have that information. If you don't see anything wrong with the data, I'd raise the soft thresholding power a bit (I assume you are going to use "signed" network type - in that case, you may want to try power 16 instead of 12), and run WGCNA. In blood methylation data, modules often relate to blood cell types; it is a bit of a sanity check to make sure you see such modules.

(duplicate post deleted, sorry)