I have problem with pairwise comparison in estimate dispersion functions in edgeR pipeline

Question

[ssko]

Hi everyone!

I have silly questions, I checked edgeR user's guide and I didn't find any answer to my question.

I have two different samples. I want to compare with each other and find differential expressions. So, I run this code : y1 <- estimateCommonDisp(y, verbose = TRUE) and it gave me this message :

Warning message: In estimateCommonDisp.DGEList(y, verbose = TRUE) : There is no replication, setting dispersion to NA.

I get the same warning when I use estimateDisp() and estimateTagwiseDisp() functions.

What does it mean 'there is no replication' because I already have two samples, how can I create the replication?
Do you have any suggestion to find the differential expression for two samples?

Thank you all!

[Princy]

Hi, If you have no replicate for your samples, then you can also use NOISeq package. The Pairwise Differential Expression Analysis (Without Replicates) functionality offers a strategy for analyzing RNA-seq datasets that do not have replicates. It is based on the software package NOISeq, which belongs to the Bioconductor project. NOISeq is a novel nonparametric approach for the identification for differentially expressed genes from RNA-Seq count data.

Answer 1

[Gordon Smyth]

See Section 2.12 "What to do if you have no replicates" in the edgeR User's Guide.

Replication means having at least two samples in at least one of your groups. Having two samples, one in each group, is not replication. Replication is a fundamental statistical concept that allows statistics programs like edgeR to estimate the amount of biological variablity that is present.

[cpad0112]

I think you meant Replication means having at least two samples in each one of your groups.

[Gordon Smyth]

No, I meant "in at least one group". It is not necessary that group sizes be balanced or that every group have replicates.