Dear Biostars:

The basic design of my RNA-seq experiment is like this:

Tissue Sample Condition

A Sample1 M5

A Sample1 P2

A Sample2 M5

A Sample2 P2

B Sample1 M5

B Sample1 P2

B Sample2 M5

B Sample2 P2

so, I'm following the paired sample design with formula = ~ Tissue+ Tissue:Sample + Tissue:Condition

My question is related to the dispersion estimated by DESeq2. I have estimated the dispersion using all fitType "mean", "local" and "parametric", but none of them seem to work shrinking the gene-wise estimates (please find enclosed the plots).

I know the dispersion fitting depends on several parameters like sample size (Tissue A= 29, Tissue B=37). Now, the question is if this is the best I can get or if there is something else to get a "better" fit.

Thanks!

fitType = Mean > http://imgur.com/uZbsDqr

fitType = local > http://imgur.com/biwXGoL

fitType = Parametric > http://imgur.com/sR0NyzR