Hello,

I am very new to this and it might be a very easy one line fix, but I have a quick question about running DESeq2 for 3 samples groups.

I performed ATACseq and RNAseq on 24 different patient samples across 3 different subtypes of leukemic cells (subtypes "E", "H", and "D".  I want to see what sites/genes are specific to each subtype using DESeq2.  I generated a Matrix with the counts file (below) and have 3 covariates being batch of sequencing (Day), sex of the patient (Sex), and the subtype (Subtype).  I then ran DESeq2 on this reducing Day and Sex and looking at Subtype dIfferences.  I then ran the results and included the Wald test to test for multiple samples.  The Wald test gave me the following table comparing "H" vs "D" (below).

I am interested in now looking at subtypes H vs E and E vs D.  Am I missing something (i.e. contrasts)?  Will contrasts work if my data is in a matrix?  Is this the best way to look at sites/genes specific to each subtype? 

> Patient_Subtype_DESeq1<-DESeqDataSetFromMatrix(countData = counts_ATAC, colData = info_ATAC, design = ~Day+Sex+Subtype)

> Patient_Subtype_DESeq2<-DESeq(Patient_Subtype_DESeq1, test = "LRT", reduced = ~Day+Sex)

> results_Patient_subtype<-results(Patient_Subtype_DESeq2, independentFiltering = FALSE, test = "Wald")

> head(results_Patient_subtype)

log2 fold change (MLE): Subtype H vs D 

Wald test p-value: Subtype H vs D 

DataFrame with 6 rows and 6 columns

Thank you for your help!

Best,

John

The contrast argument of results() lets you compare groups when there are more than two. Take a look at the ?results help file for some example usage and also the vignette.

For starters, I'd do PCA as shown in the vignette, see if sex and day really matter much.  Day of RNA extraction and day of library prep very well might matter, but if those are the same, day of sequencing run probably won't matter.  But after that, to specifically compare one group to another, use contrasts.