We're analysing some RNA-seq data using limma voom. The design of the study is very similar to https://support.bioconductor.org/p/52920/. In the past I have used blocking (subject and treatment or time point) and then used voom with duplicateCorrelation. This time in addition to subject and treatment effects we're also trying to remove the effect of a contaminating cell line. There will be varying levels of contamination but there is a very specific gene that is expressed in this cell line that can be used to quantify the amount of contamination present in each sample.
The question I have is: do we add the marker gene together with the blocking variables (i.e. part of "treat") or regress out the residuals like what is commonly done with nuisance variables in single cell RNA seq (https://satijalab.org/seurat/v3.0/cellcyclevignette.html) prior to performing DGE analysis?
Many thanks in advance. Miha