I am reading a section in the DEseq2 workflow about pairwise comparison in multiple groups Here. I was wondering how to code a simple comparison between two conditions in a multi-condition comparison. For example: if I have 3 replicates for 3 cell types (A,B,C) collected at 3 time points (0,1,3 hours) (total 27 RNA-seq datasets). How should we set the contrast to compare the replicates of A-0 hour to A-3 hours? Thanks!!!

HI,

You probably need to define a new factor that comprises a compound of Celltype and Timepoint, and then use this in your design formula.

~ Celltype_Timepoint

Then, your comparisons would look like:

A1_vs_A0 <- results(dds, contrast = c('Celltype_Timepoint', 'A1', 'A0'),
  independentFiltering = TRUE, alpha = 0.05, pAdjustMethod = 'BH', parallel = TRUE)
A1_vs_A0 <- lfcShrink(dds, contrast=c('Celltype_Timepoint', 'A1', 'A0'), res = A1_vs_A0)

A3_vs_A0 <- results(dds, contrast = c('Celltype_Timepoint', 'A3', 'A0'),
  independentFiltering = TRUE, alpha = 0.05, pAdjustMethod = 'BH', parallel = TRUE)
A3_vs_A0 <- lfcShrink(dds, contrast=c('Celltype_Timepoint', 'A3', 'A0'), res = A3_vs_A0)

B3_vs_A3 <- results(dds, contrast = c('Celltype_Timepoint', 'B3', 'A3'),
  independentFiltering = TRUE, alpha = 0.05, pAdjustMethod = 'BH', parallel = TRUE)
B3_vs_A3 <- lfcShrink(dds, contrast=c('Celltype_Timepoint', 'B3', 'A3'), res = B3_vs_A3)

There are likely ways to fit a more elaborate model, e.g. via an interaction term, but the way presented above is the easiest / most simple.

Kevin