DESeq2 paired design with missing samples
0
Entering edit mode
Geoff • 0
@geoff-24509
Last seen 9 days ago

Hi,

I've got a paired experiment with a setup like below.

The x's indicate a sample present. Each sample is a primary cell culture preparation from the same donor, and treated with one of three agonists and a vehicle control. I want to contrast each agonist independently against the control (Agonist 1 vs Control, Agonist 2 vs Control, Agonist 3 vs Control), but wondered how to address the missing data.

Reading the response on DESeq2 paired design with missing samples and multiple conditions, it seems like I have a similar situation, however, I want to confirm. My main interest is estimating the condition effects while controllign for the sample effect.

As far as I can tell, it seems like I could use the design ~Sample + condition then contrast the results object. Would this be recommended or would it be better to either subset the data or use limma-voom?

Geoff

missingdata DESeq2 paired-sample • 50 views
0
Entering edit mode
@mikelove
Last seen 2 days ago
United States

You can just use ~sample + condition. It will use available samples to estimate the sample effects and the condition effects. The coefficients are identifiable.