I am trying to setup DEseq from RangedSummarizedExperiment, and it fails because I cannot find the column for the formula.

load("data/rse_gene.Rdata")
dds <- DESeqDataSet(rse_gene, design = ~some_condition) # some_condition is a random name, it is not in the data.

It fails with:

Error in DESeqDataSet(rse_gene, design = ~some_condition) : 
  all variables in design formula must be columns in colData

With a plain counts file, I add this column after loading to the data.frame and everything works. Do I need to add some_condition to the RangedSummarizedExperiment object?

I have tried this with 3 different rse_gene.Rdata files from recount2 - all have the same issue.

Thx

The documentation says that the input has to be (emphasis mine)

a RangedSummarizedExperiment with columns of variables indicating sample information in colData

So your rse_gene needs to have that colData, and of course you pick a design based on Coldata, just like the warning says.

If for some reason you really want to make the dds object without a design, make design ~ 1.