Which fit to consider in DESeq2? Local, Parametric or Mean

Question

[dipannitas]

Hi. I am using Galaxy, a web-based platform for analysis of my RNA-Seq data. I have two types of samples, each having 2 replicates. In DESeq2 there are three options for considering the fit- Local, Parametric and Mean. How do I decide which one to go for? I am a beginner and I have looked up online, haven't got any simple answer for this yet. Looking forward. And thanks in advance!

[ATpoint]

You should keep everything at default unless you have the experience and knowledge to actually diagnose and explore what the different parameters do. There is also no simple answer to this as it goes into how the math of the method works.

Answer 1

[Mark]

There's a very good answer by Mike Love, one of the authors of DESeq2 here: https://support.bioconductor.org/p/81094/

tldr: parametric is default for a reason (it's good for most data). But local fit could potentially be better but you have to do some calculations.