DESeq LRT / 2-way ANOVA?

[sbcn]

Dear all,

Let's start from the experiment (RNA-seq) I would like to analyse:
3 time points (t0, t1, t3)
2 different parts of the brain (front, back)
2 different layers of the brain - say horizontal layers which go from front to back (layer1, layer2)

All experimental groups are present in triplicates, for a total of 36 samples.

I would like to check genes that change over time, and the influence of either the parts of the brain, or the layers of the brains, or both, on that change.

If I understand correctly, I should go for a likelihood ratio test (LRT) in DESeq2.

If I wanted to check only the influence of the brain part over time on gene expression changes, I would go for the following design:

testLRT <- DESeq(ddseq, test="LRT", full = ~ time + part + timeart, reduced = ~ time + part)

correct?

But when I want to include in that analysis the "layers", and study the possible interactions that they may have with the parts of the brain over time, it is a little more complex...

Suggestions:

testLRT <- DESeq(ddseq, test="LRT", full = ~ time + part + layer + timeart + time:layer, reduced= ~ time + part + layer)
or
testLRT <- DESeq(ddseq, test="LRT", full = ~ time + part*layer, reduced= ~ time + part + layer)

Does one of them make sense?

Thanks!

Sarah

[sbcn]

Anyone?

Hi, does someone have an opinion or some tip about this?
Thanks!