Is this the correct DESeq2 design for model with interaction terms and batch effect and paired samples?

Question

[marieke]

Hello,

I'm looking if someone can confirm or adjust the design for DESeq2 for an experiment with an interaction term (Genotype and Treatment), a batch effect (Differentiation_replicate) and paired samples (DonorID).

Based on the DESeq2 Vignette, these are the three recommended adjustments:

• Interaction term: make a new column with the two columns that interact into 'group' and use design = ~ group
• Batch effect: design = ~ batch + condition
• Paired design: design = ~ subject + condition

Is it correct to combine these recommendation into the following model?

design = ~ batch + subject + group

or in our case

design = ~ Differentiation_replicate + DonorID + Genotype_Treatment

Thanks a lot!

[ATpoint]

Can you post colData(dds)? That will help getting an idea how donors relate to the design here.

[marieke]

Sorry for my late answer! This was more of a theoretical question. I'm analyzing multiple projects, and some of them need a batch effect correction and some have paired samples, and I was wondering which design would be used if I need to analyze a project that requires both. I currently don't have a project like that so I don't have any coldata(dds) to share.
I've now also realized that it would depend how the donors are distributed over the batches, so I'll guess I'll come back here with more info if I encounter that situation. Thank you anyway!

[ATpoint]

Yes, batch and donor are usually not compatible, do one or the other. Donor (so paired analysis) is imo always preferred as most powerful, generally speaking.