Question: DESeq2 3 factor design - whether to split up data or add interaction terms
gravatar for atongz652
10 weeks ago by
atongz6520 wrote:

I have 60 birds in a study, we split up the birds so that there are 15 birds per treatment (so 4 treatments in total), and at 3 different time points, 5 birds were sacrificed Each bird contributed 6 different sample types(different areas of the body were sampled). I want to study the influence of treatments (ie. contrast different treatments), but I already know that both sample type and time point a major impact on gene expression expression.

For example, I want to know at timepoint 0, sample type 1, which genes are deferentially expressed between treatment 1 and treatment 2, or between treatment 2 and treatment 4, etc.

Should I split the data up into 18 smaller datasets so that each smaller dataset only contains samples of the same sample type and time point, then use design= ~treatment?

Alterantively would design=~Site+Timepoint+Treatment on my entire dataset tell me give me a good idea of genes that are differentially expressed after accounting for site and timepoint? Are there interaction terms I need to take into account? Thank you!

deseq2 design • 80 views
ADD COMMENTlink modified 10 weeks ago by Michael Love22k • written 10 weeks ago by atongz6520
Answer: DESeq2 3 factor design - whether to split up data or add interaction terms
gravatar for Michael Love
10 weeks ago by
Michael Love22k
United States
Michael Love22k wrote:

The question of whether and which interactions to include when you have three factors is beyond the amount of support I can give here. It’s more of a statistical consulting question than a software question and I don’t have the extra time these days to get into it.

Any of these models can be set up in DESeq2 and coefficients or combinations of coefficients pulled out, similarly to a linear model in R.

Rather than split into separate datasets, for simply comparing levels of treatment within site and timepoint, see the Interaction section of the vignette. We have a relevant suggestion there.

ADD COMMENTlink written 10 weeks ago by Michael Love22k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 111 users visited in the last hour