Hi all,

this could be quite straightforward so bear with me!

I have the following sampleTable

    sampleTable
                 sample type
Sample_1_S1 Sample_1_S1    A
Sample_2_S2 Sample_2_S2    A
Sample_3_S3 Sample_3_S3    A
Sample_4_S4 Sample_4_S4    B
Sample_5_S5 Sample_5_S5    B
Sample_6_S6 Sample_6_S6    B
Sample_7_S7 Sample_7_S7    C
Sample_8_S8 Sample_8_S8    C
Sample_9_S9 Sample_9_S9    C

with which I have already done type contrasts with

res_AB <- results(dds, alpha = 0.05, filterFun = ihw, contrast = c("type", "A", "B"))

I would like to study up-regulated genes per type. For example, type "A" alone

sample type
    Sample_1_S1 Sample_1_S1    A
    Sample_2_S2 Sample_2_S2    A
    Sample_3_S3 Sample_3_S3    A

however, I am unsure what is the best way to do this. So far, I assume multiple contrasts per type A with 3 sample levels would work

res_A12 <- results(dds, alpha = 0.05, filterFun = ihw, contrast = c("sample", "Sample_1_S1", "Sample_2_S2"))
res_A13 <- results(dds, alpha = 0.05, filterFun = ihw, contrast = c("sample", "Sample_1_S1", "Sample_3_S3"))

and I sort for most highly expressed genes shared across samples from type A that way but is there a better approach to this? I would like to group samples to type and make individual type-specific comparisons (within type, A, B, and C, respectively) that way but I am unsure if this will fit to DESeq2's model.

Thank you for your time!

Shauna

If you wanted to be up regulated in A compared to B and compared to C, you can just build a set that is the intersection of the A vs B and A vs C comparison. This is my preferred approach rather than A vs the average of B and C which can be misleading when they go in different directions.