DESeq2 -- Three conditions with "gradient"

Question

[mschmid]

I am analysing data which I have mapped to the reference genome using HISAT2. I estimated the expression using Stringtie.

Now I want to run DESeq2 on this data. Works well so far. But I am not sure how I have to "model" my experimental design.

I have:

• 5 x biol. replicates "Control (no treatment)"
• 5 x biol. replicates "Treatment low dosage (chemical compound)"
• 5 x biol. replicates "Treatment high dosage (chemical compound)"

Do I have to compare control<>low dosage and control<>high dosage individually or is there a better way to model the "gradient" of treatment using DESEq2?

Answer 1

[swbarnes2]

You can do a dose response model, though you might want to play around with the number values you use to represent zero, high and low treatment.

The results for this will say they are returning a log fold change, but they are not; they will instead be returning the slope of the line, and the p-value is how well the points line up on the line.

https://support.bioconductor.org/p/88254/