DESeq2 no replicates analysis with a design value of ~1 generates only positive log2foldchange

Question

[BIN]

Hello, I am trying to understand the output of my DESeq2 analysis when running it without any replicates. I know this type of statistical analysis is known to be useless when you have no replicates, but nonetheless, I'm still curious on the DEG that will show up from this. Since the new version of DESeq2 doesn't let you do an analysis without replicates I found a way to still make it run, but by using the design method of ~1 ( intersect only ). This allows the package to run without any fatal error for my data. However the log2foldChange that are generated are only positive values... When I plot my result ( before normalisation ) this is the volcano plot generated...

Can anyone help me understand this please ? Thank you in advance for your time!

Answer 1

[ATpoint]

This is a nonsense way of running the analysis which has no practical meaning in terms of differential expression, see from the DESeq2 author: https://support.bioconductor.org/p/79209/

DESeq2 needs replicates by design. If you have none, and want to do some kind of exploratory analysis then please see the edgeR user manual. It has a section on no-replicate situations. Basically you would make up a dispersion estimate which then pretends you have replicates. Of course this is neither robust nor publishable but at least you can get some very preliminary idea of what is going on in your dataset. Keep in mind that for a publication or any confident statement you will need to repeat the experiment witha properly replicated design.

[BIN]

Alright thank you for your time, I will look into edgeR !