Hi there, I am using DESeq2 to analyse RNAseq from mutants and control samples. I have 3 controls and 2 mutant samples from distinct individuals but in addition I have a sample where the mutation has been corrected and another where the mutation has been inserted. So in total I have 4 samples without the mutation and 3 samples with the mutation:

mutation   cellline  original_individual isogenic          
  <fct>    <fct>     <fct>               <chr>             
1 ctrl     ctrl1     ctrl1               no                
2 ctrl     ctrl2     ctrl2               no                
3 ctrl     ctrl3     ctrl3               no                
4 ctrl     isomut2   mut2                mutation corrected
5 mut      mut1      mut1                no                
6 mut      mut2      mut2                no                
7 mut      isoctrl3  ctrl3               mutation inserted

I would like to contrast mutation_mut_vs_ctrl but remove effects due to the original_individual to get a clearer picture of differences attributable to the mutation.

In order to control for differences due to the original_individual, I have specified the design: ~original_individual + mutation and the contrast mutation_mut_vs_ctrl. Is this optimal or will accounting for the original_individual confound the mutation comparison?

For example, would it be preferable to specify original_individual:mutation as an interaction design (this gives error: model matrix is not full rank).

Many thanks! Oliver

For discussion of statistical design and interpretation of results I recommend working with a local statistician or someone familiar with linear models in R. I have to save my time on the support site for software related questions.