I face some problem when I apply WGCNA in the RNseq data. Here’s the thing,

I have the raw court data, which looks like as:

head(RawCount_all_Filter)

                P6_A1 P6_A2 P6_A3 P6_A4 P6_A5 P6_A6 P6_A7 P6_A8 P6_A9 P6_A10 P6_A11 P6_A12 P6_B1 P6_B2 P6_B3 P6_B4 P6_B5 P6_B6 P6_B7 P6_B8

ENSG00000000419     0   375   527   337     0   607   595   149     0      0      0      0     0     0   567     0   322     1     0    42

ENSG00000000457     0     0     0     0     0     0     0     0     0      0      0      0     0     0     0     0     1     0   283     0

Then, I directly apply the count data to vsd = varianceStabilizingTransformation(RawCount_all_Filter), which looks fine.

I got the Expr = as.data.frame(t(vsd)) and Bench = t(Phenotype_all).

Then, checking the data, and cluster the sample as usual.

https://bri.box.com/s/solx4maup2g46jbcg1qe14s9vnszzsok

The height looks not like as usual in the microarray data, which I don’t if that’s correct.

But the scale free topology figure looks fine:

https://bri.box.com/s/9tn9frq2asajqgk48haf806av4j0vkk6

I select power = 10.

Then calculate the Co-expression similarity and adjacency matrix and TOM,

softPower = beta1

adjacency = adjacency(Expr, power = softPower,type = "signed")

TOM = TOMsimilarity(adjacency,TOMType = "signed")

dissTOM = 1-TOM

I get modules like this:

https://bri.box.com/s/35br575uafusinti45a0e9nuywghn8lk

> table(dynamicMods)

dynamicMods

   0    1    2 

  14 8041  305 

I pretty shock about the results, and I don’t know how to interrupt or modify the analysis.

Later around, the module-trait relationships look awful,

https://bri.box.com/s/ofawxu03hlibrcik3totja66qn6s40z1

Could you please point out what’s going on my data? And how to improve it?

Thank you so much!

It is impossible to diagnose issues with your data or analysis without seeing the data (the code looks reasonable but it obviously isn't complete). One usually sees at least a few distinct branches in WGCNA analysis of expression data, but again, without knowing what your data represent and seeing the actual data it is impossible to say anything of substance.