Entering edit mode
10 months ago
Ömür Koray
•
0
Hello,
As a newbie I was trying to do WGCNA for my transcriptomics data. I used vst normalization and tried to choose soft power. Even in high numbers I couldn't reach 0.8 threshold. Is there any suggestions for overcome this problem?
Have you performed any sample filtering or gene filtering? I would (1) check the top few PCs for outlier samples, and exclude; (2) Proceed with WGCNA at a moderate power (12-16) and inspect the dendrogram, looking for tiny gene clusters (2-4 genes) with very strong/outlier linkage, and investigate whether those should be retained.
If you have a very strong gene expression driver (say you're looking at cell lines or treatments), then you may prefer to use truncated R2 in place of R2.
Thanks for your reply. I have tried without filtering & filtering out low counts with (rowMeans > 0, 5, 10) but soft power threshold graph remains kind of same. Regarding to your first approach since I have low number of samples (6) and i just have age variable between my samples, no treatment. I am also attaching my PCA plot.
n=6 is very low for WGCNA. The general advice would be to set softPower to something like 12-16 and see what comes out.