Search
Question: DEseq2 - minimum number of genes
0
gravatar for stefanie.graindorge
6 days ago by
stefanie.graindorge0 wrote:

Hi all,

I will have to perform a differential expression analysis on RNA-seq data obtained from purified chloroplasts.

That means that the mapping will be done on the chloroplastic genome (around 80 genes) and so the DE analysis.

I was wondering if there is a minimum number of genes to be considered to have a correct DESeq2 analysis. And if there are specific parameters to change to do this.

Thanks a lot for your help

Stefanie

 

ADD COMMENTlink modified 6 days ago by Michael Love19k • written 6 days ago by stefanie.graindorge0
2
gravatar for Michael Love
6 days ago by
Michael Love19k
United States
Michael Love19k wrote:

There isn’t a modification, the prior will be learned over the 80 genes. I think the dispersion estimation will be fine. These days I’m more concerned about size factor calculation in situations like this. Can you assume a peak of genes (in a histogram for example) will have LFC near 0?

ADD COMMENTlink written 6 days ago by Michael Love19k

I can't say for sure. There will probably be some genes with LFC around 0. It is a new experiment and we have no clue about how these genes are going to behave between the two conditions... Do you think we should add some "fake" counts with a LFC of 0 to better fit the null hypothesis of DEseq2 ?

ADD REPLYlink written 6 days ago by stefanie.graindorge0

Adding fake data isn't going to solve anything, it's just going to make the data worse. If you do so, you'll effectively be making an arbitrary determination of the size factors with no evidence while also disrupting the dispersion estimation.

ADD REPLYlink written 6 days ago by Ryan C. Thompson6.9k

Oh so bad idea :(

So in summary, I will check if there is a peak of genes around LFC 0 and if this is the case I can assume the analysis will be ok. And if not I will have to see how to normalise the data differently to compare the samples.

Thanks a lot for your quick answers.

 

ADD REPLYlink written 6 days ago by stefanie.graindorge0

You can’t tell from the data if global scaling is inappropriate.

You’d need to know some biology about the genes.

ADD REPLYlink written 5 days ago by Michael Love19k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 437 users visited in the last hour