DeSeq-1 and edgeR

Question

[xd_d]

Hello,

i have a question to the exact test with the negative binomial distribition to identify differently expressed genes.

I dont find a good explanation of these test. Not in the papers of edgeR or DeSeq-1(2010)

Only explantion is: similar to Fisher-exact test. But thats say nothing...

Have some one a good source for these exact test ?

[Arup Ghosh]

The DESeq(1) paper:

https://genomebiology.biomedcentral.com/articles/10.1186/gb-2010-11-10-r106

Answer 1

[Kevin Blighe]

They do not give it any particular name. For DESeq, it was just referred to as DESeq. They help users to understand it better by stating that it is akin to a Fisher's Exact test. It is essentially it's own statistical test developed for the purpose of determining probabilities from RNA-seq. The formulae used are in the manuscript whose URL arup has provided.

Some things to remember:

1. DESeq / DESeq2 both model data as a negative binomial, in which the mean-to-variance relationship can vary. Earlier methods modelled the data as a Poisson, in which the mean=variance.
2. In DESeq2, the statistical test was switched to the Wald Test, which is applied on a gene-wise basis on the fitted negative binomial model.

Some other related posts:

• What statistical test does DESeq use?
• DESeq statistical test
• [DESeq2] Changes compared to DESeq
• Differential Expression analysis using Wald-Test
• A: Clarification on how DSEeq2 Dispersion Curve is Generated