Hello,

I'm running DESeq with the design=~Cell_Treat + MGMT.  I'm working with 6 samples from two different patients, so 12 samples in total.  There are two factors, Cell line (HF2303 or HF 2927) and Treatment (untrt or trt).  I'm trying to include a continuous covariate into the design (MGMT) to control for the effect of the factors based off of MGMT presence in the cell.  

 > colData(se)
DataFrame with 12 rows and 5 columns
                            sample     Cell     Treat     MGMT Cell_Treat
                          <factor> <factor> <integer> <factor>   <factor>
1   accepted_hits.CSC.Contc.HF2303        A         1      pos        A-1
2    accepted_hits.CSC.TMZa.HF2303        A         2      pos        A-2
3   accepted_hits.SDC.Contb.HF2303        A         1      pos        A-1
4   accepted_hits.SDC.Contc.HF2303        A         1      pos        A-1
5    accepted_hits.SDC.TMZb.HF2303        A         2      pos        A-2
...                            ...      ...       ...      ...        ...
8    accepted_hits.CSC.TMZb.HF2927        B         2      neg        B-2
9    accepted_hits.CSC.TMZc.HF2927        B         2      neg        B-2
10  accepted_hits.SDC.Contc.HF2927        B         1      neg        B-1
11   accepted_hits.SDC.TMZa.HF2927        B         2      neg        B-2
12   accepted_hits.SDC.TMZc.HF2927        B         2      neg        B-2

When I try to set up the DESeqDataSet with design=~Cell_Treat + MGMT it gives me an error that the model matrix isn't in full rank.  I thought this might be due to that MGMT factors correlate the same as the Cell factor, but I'm not sure if I should get rid of the column because id like to base the results off of whether or not the MGMT variable is present of not and the cell column would be needed to separate HF2303 from HF2927?  

Am I doing something wrong in the way the design is set up or columns?  

This is the full error message:

> dds<-DESeqDataSet(se, design=~Cell_Treat + MGMT)
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')

You say that MGMT is a continuous variable, but in your DataFrame it is coded as a factor with two levels, so you should clarify which one it is. If MGMT is a factor that is dependent on the cell type, then it is completely confounded by cell type, and cannot be in the model at the same time as cell type. In other words, both of those factors tell us the same exact thing, so you can't solve for them as if they are different things.
 

Let me just add a quick suggestion that you discuss your desired analysis with a local statistician at your institute. 

They can help explain in person the issue with including MGMT and Cell_Treat at the same time in the model, and so help guide your choice of design and the interpretation of results.

While it's possible to run statistical software without a statistical background, a lot of the basics and interpretation of the results can best be explained in a face to face conversation.

While we've added many automatic warnings and error checks to uncover modeling problems, the lack of an error or warning doesn't necessarily mean that you're getting the analysis you want.