Hi everybody,

I am wondering about building the right design/contrasts for addressing the question I have to my data.

My DESeq2 gene expression data set is from cells which can have two differentiation states S0 and S1. The cells are from humans treated with a kind of diet and the cells are extracted from blood before T0 and after T1 therapy. At T0 the cell states can be either S0 or S1 and at T1 the cell states also can be S0 or S1. Due to the therapy it happened that either cells stayed the same differentiation state or they changed the state from S0 to S1 or from S1 to S0 due to therapy. Now, I want to get the differential gene expression that occurs in S0_T0 to S1_T1 due to therapy and in S1_T0 to S0_T1 due to therapy. I want this isolated from the gene expression differences of S0_T0 and S1_T0 and S0_T1 and S1_T1, respectively. How can I work this out?

ChatGPT said I should use the design: ~ differentiation_state + Therapy_T + differentiation_state:Therapy, however this gives me the contrasts

[1] "Intercept" "differentiation_state_S0_vs_S1" "Therapy_T1_vs_T0"
[4] "differentiation_stateS0.TherapyT1"

And I am not sure if this is what I want.

Any help would be very much appreciated!