Hi,

I am analyzing NanoString gene expression data with DESeq2. The aim is to identity differently expressed genes between advanced and not advanced renal cancer ("status"). The measurements were carried out in three cartridges, and unfortunately all of the samples with inferior RNA-quality (low RNA-content) were measured in the third cartridge. (Fortunately,  the advanced cancer samples are not significantly over/under-represented in the third cartridge.)

Question: Should the cartridge be included as a confounding variable in the DESeq2 analysis (which model should I use: design= ~ cartridge + status, or design= ~ status)?

Yes I would typically add any such batch effect to the design. You can use plotPCA to see the batch effect by cartridge (see vignette).