Hi there,

I am trying to run a DE analysis with DESeq2 on an in vivo RNAseq dataset, but am having trouble accounting or the control group when comparing the treatment groups. I have one DMSO-treated control group plus 2 drug-treated groups, both treated with the same drug, but one group having been sequenced while tumors where in remission (group 1), and one group having been sequenced when tumors had become resistant and had been growing out again (group 2). As below (I have more replicates for each group, but it essentially looks like below).

Sample - Group

DMSO_1 - DMSO

DMSO_2 - DMSO

GroupA_1 - A

GroupA_2 - A

GroupB_1 - B

GroupB_2 - B

I initially defined the DESeq2 design and contrasts as

# Create DESeq2 object
dds_object <- DESeqDataSetFromMatrix(countData = data, colData = metadata, design = ~ Group)`

# Define contrasts
dds_res_1 <- results(DEanalysis_invivo, contrast = c("Group", "group_1", "DMSO"), alpha = 0.05)
dds_res_2 <- results(DEanalysis_invivo, contrast = c("Group", "group_2", "DMSO"), alpha = 0.05)
dds_res_3 <- results(DEanalysis_invivo, contrast = c("Group", "group_2", "group_1"),alpha = 0.05)

Now I am interested in comparing the DE genes between the resistant and acute groups, but I am unsure how to account for DMSO in that comparison. I have looked at the DESeq2 vignette and added a column indicating Condition (drug or control), and added that to the design. But looking like this, there is a linear dependency of Group and Condition, since they basically say the same thing.

Sample - Group - Condition

DMSO_1 - DMSO - control

DMSO_2 - DMSO - control

GroupA_1 - A - drug

GroupA_2 - A - drug

GroupB_1 - B - drug

GroupB_2 - B - drug

So when I run the DESeq2 object with the design as

dds_object <- DESeqDataSetFromMatrix(countData = data, colData = metadata, design = ~ Group + Condition)

it gives me an error since there is a linear dependency between the two and one needs to be removed. I guess it essentially requires the data to have a control equivalent for both group 1 and group 2, yet the issue with this of course is that group 1 and 2 are not treated with a different drug or dose, they are essentially different time points and in the setup of this experiment, they share the same control group (DMSO).

So because I am interested in the DE genes between the resistant and acute, my question is how do I account and correct for DMSO in the comparison between those? And if I cannot do that in the DESeq2 design, are the DE results between group 2 and 1 reliable, and what other precautions would I need to take?

Any help would be appreciated, thanks in advance

Now I am interested in comparing the DE genes between the resistant and acute groups, but I am unsure how to account for DMSO in that comparison.

This has been asked before. If you think of it in terms of algebra: you have say A, B, C. You don't need A to compute C - B.

You can search the site for similar posts, but it's just not relevant for that pairwise comparison.