Just to complete the tutorial, use a small dataset. For your actual analyses, especially if you will integrate multiple samples/datasets, you will probably need access to an HPC.

The model that is fit to the counts data is a negative binomial generalised linear model which is still a linear model.

As you should usually, unless a:b is very small and not significant, then just use a+b. Simpler equations are subjectively better. You can use "a*b" as shorthand for "a + b + a:b". It makes it a lot shorter when you have an experiment like "a*b*c", short for "a+b+c+a:b+a:c+b:c+a:b:c"

@@qwerty11111122 Damn I didn't know that. Thanks a lot!

Yes, but put "genotype" at the beginning. And also if you are interested in interaction terms, it would be: design = ~ genotype + treatment1 + treatment2 + treatment3 + genotype:treatment1 + genotype:treatment2 + genotype:treatment3 If you are applying each treatment separately though, meaning that your design matrix looks something like this: genotype treatment sample1 I NONE sample2 II treatment1 sample3 I treatment2 sample4 II treatment3 then you just do genotype+treatment+genotype:treatment as shown in the video.