Multiple variables/interactions

[mtruglio]

Description of feature

Hello,
My experimental design is longitudinal (i.e. time series), and my main focus is finding DE genes in the interaction between treatment (yes/control) and time (T0....T6). My metadata header looks like this:

sample treatment time

The DESeq2 model would be something like:

dds <- DESeqDataSetFromMatrix(countData = counts,
                              colData = colData,
                              design = ~ treatment + time + treatment:time)

Now, as far as I know, the contrast file only allows one variable and one reference level. Is there a way to use this pipeline for my case (two variables, two reference levels) with some contrasts magic? If not, can it be implemented?

Thanks

[pinin4fjords]

So, we haven't yet engineered any functionality for this explicitly as you have noticed. PRs welcome- the DESeq module is here: https://github.com/nf-core/modules/tree/master/modules/nf-core/deseq2/differential

I haven't though hard about this but.... if you included the extra parts in the blocking column like time;treatment:time it would get into the model, if it were not for the checks carried out here. You could try removing that check on a fork and seeing if that worked.

[seb-mueller]

Same issue here, any updated on this issue? Does the suggested solution work?