Hi, My experiment is composed by 3 cell fractions: cytosolic, nuclear and total and 3 groups: siRNA, siRNAscramble and nontreated. Both siRNAscarmble and nontreated are part of the controls group.
I want to test for differentially expressed genes by compare the same fraction (e.g nuclear vs nuclear) of siRNA vs controls by accounting ( remove ) for differences among siRNAscramble and controls.

How can I write the DESeq2 design formula?

Thanks,

Choice of the design formula is really a statistical consultation. I unfortunately have to restrict my time on the support site to software related questions, and recommend that users who are new to linear model discuss with a local statistician or someone familiar with linear modeling in R.

Look in the vignette for interactions. That's what you want. You don't think about it as literally subtracting out things, you think about it as looking for genes which change differently in response to the treatment depending on the background condition.