How to do DAP analysis with multiple controls and multiple treatments

[danielcgingerich]

Hello,

I have 10 control samples and 10 treatment samples. I want to do DAP analysis between control and treatment.

I see a few different ways of approaching this:

1. RunTFIDF on all 20 samples separately, then pool control and pool treatment for DAP analysis
2. Merge control datasets, RunTFIDF on merged data // merge treatment datasets, RunTFIDF on merged data. Then do DAP analysis
3. Integrate controls, Integrate treatment, then do DAP analysis. I realize that integration assumes dependencies between control and treatment, but in this case, I am integrating control with control, and treatment with treatment, thus assuming no dependencies across conditions

[timoast]

I either 1 or 2 seem fine to me, you could try both ways and compare them. If you include batch as a latent variable in the differential accessibility model, I wouldn't expect there to be much of a difference in the results. I would not do option 3.