Dear all,

I am performing a differential expressed genes analysis using DESeq2 and I want to correct the design by donor and by season (winter vs. summer where the extraction was done). The problem is that the the donor corresponds with the season in sets.

An example of data is:

         id time season
sp1  ID_003    0      W
sp2  ID_003    4      W
sp3  ID_003    6      W
sp4  ID_003    8      W
sp5  ID_004    0      W
sp6  ID_004    4      W
sp7  ID_004    6      W
sp8  ID_004    8      W
sp9  ID_005    0      S
sp10 ID_005    4      S
sp11 ID_005    6      S
sp12 ID_005    8      S
[...]

For hence, when I add the design ~ id + time*season, DESeq returns the following error:

R> dds.blood <- DESeqDataSetFromMatrix(
+   countData = counts,
+   colData   = desc,
+   design    = ~ id + plate_batch + flowcell + time + season + time:season
+ )
Error in checkFullRank(modelMatrix) :
  the model matrix is not full rank, so the model cannot be fit as specified.
  One or more variables or interaction terms in the design formula are linear
  combinations of the others and must be removed.
See the section 'Model matrix not full rank' in vignette('DESeq2')

So my questions is: There is a right way to correct by donor in this case were season corresponds in block with the donor? I'm using DESeq2_1.8.1

As the error suggest I read the section Model matrix not full rank from DESeq2's vignette and I followed the instruction:

<<In both of these cases, the batch effect cannot be fit and must be removed from the model formula.>>

So I removed the id from the model. But I want to be sure this is the correct way to follow up in my case.

hi,

You should try the technique which is in the vignette below the text "Finally, there is the case of an experiment with grouped individuals, where a group-specific effect of some treatment is sought". You'll want to redefine a new column where id is nested within season, and then use a design ~ season:idnested + time*season. This will control for id and allow you to test the interaction. For designs with interactions, I recommend now to set betaPrior=FALSE in DESeq().

It's also described here: A: DESeq2 paired multifactor test