DESeq2 Error in checkFullRank(modelMatrix)?

Question

[Mainul]

I have 2 factors in the design formula in DESeq2

and my colData

Con = c (rep("CT_0",3),rep("CT_2",3),rep("CT_4",3), rep("S_0",3),rep("S_2",3),rep("S_4",3), rep("CS_0",3),rep("CS_2",3),rep("CS_4",3), rep("T_0",3),rep("T_2",3),rep("T_4",3), rep("T_0",3),rep("T_2",3),rep("T_4",3), rep("S_0",3),rep("S_2",3),rep("S_4",3))

So my control total 18 (where's tolerant control 9 (CT) and susceptible control 9(CS)), Tolerant sample 18 and susceptible sample 18.

point = c(rep("C",9), rep("S",9), rep("C",9), rep("T",18), rep("S",9))

So When set design with con+ point its shows error

dds1 = DESeqDataSetFromMatrix(countData = countData, colData = nData, design = ~Con+point) dds1=DESeq(dds1)

Error

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. Please read the vignette section 'Model matrix not full rank': vignette('DESeq2')

Any expert opinion regarding this how can I design my analysis part? Every comment will be much appreciated. Thanks in advanced

Answer 1

[swbarnes2]

You can't put Con and point into the design together. Pick one, or the other, depending on what question you want to ask.

Answer 2

[Santosh Anand]

Your design is a nested design, in which the variable Con is nested in the variable point (or to say in the mathematical term that point is a linear combination of Con). See the answer by "Michael Love" in the link below to get a better idea of how this is nested, and what problems it creates in the linear modeling (used by DESeq2)

https://support.bioconductor.org/p/58553/

So, you need to drop one of the variables from the design formula. See Model matrix not full rank at

http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html

In some cases, you can calculate the effects even if you have one variable that is completely nested inside another (but you will need at least one more variable which is not linearly dependent on them). See Group-specific condition effects, individuals nested within groups in the above DESeq link to better understand this situation.