Hi,

I am using DESeq2 to identify DE genes across two color types. I have two columns – color and sample ID. As you can see, my samples are unpaired:

According to the vignette (3.12 “Model matrix not full rank”) and other questions here, am I correct that I can only perform design = ~ Color, and cannot fit the effect for Sample_ID? There is a strong sample effect in my data set, but I am afraid that this data set does not allow me to correct for such effect.

Many thanks in advance,

Miyako

You shouldn't (and cannot) control for the difference between samples in this experiment. The differences between samples inform the statistical model about the within-group variability, and you want to find differences across color that rise above the within-group variability. You should use a design ~color. If you cannot find any differences at a given FDR threshold, it means that either there is no effect or the effect is too small to detect with your sample size and you need to generate more samples.