Hi all,

I would like to use DESeq2 for the analysis of a pooled shRNA screen to identify mutation specific lethals. The design of the experiment is a number of mutant cell lines and a number of wildtype cell lines. For every cell line there is a t0 measurement and t1 measurement. Interesting hairpins/genes are those that give significant more killing in the mutant cell lines, compare to the wt cell lines. Because each cell line has it's own t0, the data need to be paired wise analysed.

I tried to do it with DESeq2 in the following design:
formula(~ line + time + condition)

A simple example:
d<-data.frame(w_l1_t0=c(4,8),w_l2_t0=c(5,9),w_l1_t1=c(4,8),w_l2_t1=c(5,9),m_l3_t0=c(4,8),m_l4_t0=c(5,9),m_l3_t1=c(1,2), m_l4_t1=c(1,2))   
row.names(d) <- c("hp1","hp2")

countData <- as.matrix(d)   
colData = data.frame(condition=c("w","w","w","w","m","m","m","m"),
        time=c("0","0","1","1","0","0","1","1"),
        line=c("1","2","1","2","3","4","3","4"))

dds <- DESeqDataSetFromMatrix(countData = countData,
    colData = colData,
    design = ~ line + time + condition)

However I then get the error:
Error in DESeqDataSet(se, design = design, ignoreRank) :
  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

I assume this is because cell line and condition are linked. Is there a way to do the above analysis in DESeq2? I don't know the exact meaning of the parameter ignoreRank, but would setting it to TRUE be an option?

Thank you in advance for your reply!

Best regards,

Cor Lieftink

P.s.  I thought it was solved, after using an extended dataset with 15 mt and 15 wt cell lines the error was gone. However I had there numbered the lines within the subset of wildtype and mutant, which should be (I think) continuous over the two sets. The error than re-emerge.

hi Cor,

You're right, the issue is that the lines are nested within condition. Here's a similar post which I gave a solution for: A: DESeq2 paired multifactor test

essentially, you should do the same steps of creating a vector "nested.line", which just distinguishes the lines within a condition. So for the above, it would be c("1","2","1","2","1","2","1","2").

Then, I think you want to use a design of ~ condition + time + condition:nested.line + condition:time. This will control for the line specific effects. And it sounds like you want to test the condition:time effect. This tests if the mutant cell lines experience a change at time 1 compared to time 0 which is different than the one for wildtype. you can use the following call:

dds = DESeq(dds, modelMatrixType="standard")
results(dds, name="conditionm.time1")

(the modelMatrixType part is only necessary so that the interaction can be tested by a single coefficient, even if you add more cell lines.)