What is the criteria of logCPM? rpkm() error

Question

[kanika.151]

Hello,

After getting results from edgeR, How do I know which logCPM values are significant and which are not?

I know that we can calculate RPKM values through edgeR by using

gdrpkm<- rpkm(data,gene.length=vector)
Error in gene.length/1000 : non-numeric argument to binary operator

[Devon Ryan]

Define "significant". logCPM values are logCPM values, there's no concept of significance typically attached to them.

[kanika.151]

My RNA Seq data does not have technical replicates as the data was pooled into one. That is no replicates which won't give conclusive results per say which is why I cannot take P value or FDR values into consideration. I wanted to know if I had to deduce from logCPM values which genes are differentially expressed and which are not how would I do that? As in which is significant and which is not?

Answer 1

[Devon Ryan]

You can ignore the logCPMs, they're not useful for you. Pretty much the best you can do with your dataset is to use GFOLD.

[kanika.151]

Ohk. hmm. How reliable is GFOLD? Are there any publications made using that software?

[Devon Ryan]

For unreplicated data it's about as reliable as one can hope. You might be able to find publications using it, though if I were the reviewer I'd reject all studies lacking replicates (I'm not alone in this).