Finding DEGenes taking into account the DEG of a compared control condition
1
0
Entering edit mode
18 months ago
mcalleja ▴ 110

Hi. I have the following problem:

The wet lab has prepared four samples (with one replicate). (It's not a paired problem)

Control phase 1, control phase 2 Treatment phase 1, treatment phase 2

We are interested in finding differential sgRNAs between treatment phase 1 and treatment phase 2. BUT we want to include the effect that derives from the absence of difference between control phase 1 vs control phase 2. But we still want to capture and see where there's (or not) a differential count in the control treatments.

Let me elaborate, a true control for us is precisely the difference (or effect) between control phase 1 vs control phase 2, is not the mean between the two, but rather if there's a DE between control phases. We would like to use this difference (control phase 1-vs control phase 2) as an effect, or to factor in the magnitude of the effect , but I am not sure how to include it. And what tool would be more appropriate.

I think want I want could be done with Limma or edgeR or anything that takes in a contrast matrix. However, a part from being able to contrast all groups between them, I don't know how to factor in the effect of the comparison between two groups, or how to make that comparison to be the common baseline.

I've came up with the following contrast matrix, but I am not sure it's correct:

baseline  Treatmentphase 1 Treatmentphase 2 Controlphase 1 Controlphase 2
Library A_1  1  1  0  0  0
Library A_2  1  1  0  0  0
Library A_3  1  0  1  0  0
Library A_4  1  0  1  0  0
Library WA_1 1  0  0  1  0
Library WA_2 1  0  0  1  0
Library WA_3 1  0  0  0  1
Library WA_4 1  0  0  0  1


So, is there a way of doing what we want nice and neat within the Contrast matrix? or is it necessary to do all the contrast and then work around (like in a VennDiagram) which sgRNA are equal and which different in control phase 1 vs 2 and then, see how they behave in treatments?

All your suggestions are more than Welcome

RNA-Seq R sgRNA limma edgeR • 341 views
2
Entering edit mode
18 months ago
mcalleja ▴ 110

I've found the answer I was looking for, after reading the Limma's User guide

Section 9.5 Interaction Models: 2 × 2 Factorial Designs 9.5.1 Questions of Interest

The experiment that is described there is exactly what we have. The basic design is precisely that one. I just need to see if I can apply it and ask factorial questions.

I'll try to workout the experiments using this factor design.