DEseq2: Advice for model (double interaction ?)

Question

[Picasa]

Hi,

I have that kind of data:

| Group | Donor | Condition | Category |
|-------|-------|-----------|----------|
| X     | 1     | KO        | Black    |
| X     | 1     | WT        | Black    |
| X     | 2     | KO        | Red      |
| X     | 2     | WT        | Red      |
| X     | ..    | ..        | ..       |
| Y     | 3     | KO        | Black    |
| Y     | 3     | WT        | Black    |
| Y     | 4     | KO        | Red      |
| Y     | 4     | KO        | Red      |
| Y     | ..    | ..        | ..       |

This is paired sample and want I to test the effect Category but for each Condition and Group so:

Red vs Black for X and KO
Red vs Black for X and WT

Red vs Black for Y and KO
Red vs Black for Y and WT

Any advice ? Thanks for your help

Answer 1

[swbarnes2]

You probably will have to drop donor as an element of your design, it's obviously nested with other things, and with just two samples per donor, I'm not sure you can model it well anyway. Simpler to just omit it.

The simplest way to compare one subset of sample to another is to make a new column of group_condition_category, make that your design, and use contrasts to specify what gets compared to what.

[Picasa]

Thanks for your advice. I will try this