I'm doing WGCNA co-expression analysis on 29 samples related to a specific disease, with RNA-seq data with 100million reads. After all quality control, I ended up with 53000 genes in FPM measure. I then filtered all genes with 0 variance, miRNA and pseudogenes, also genes that had less than 10 reads across more than 90% of samples. I ended up with 14000 genes.
When I run goodSamplesGenes, everything seems ok. I run power treshold and ended up choosing 7 (graphs attatched) Then I run WGCNA blockWiseModules(), with block size 5000 (tried 7000 and 8000 with no good results). I will do an analysis with the full data with our HPC.
But I got one big turquoise module, which is very unexpected.
If I change the network to "signed hybrid", I start getting a little bit more clusters: