I am struggling to get my head around the correct design to use for my experiment. I have RNAseq of 2 pure species (s1, s2) and of the hybrid (h). I aligned all three genotypes to the s1 genomes. I am interested in the comparison of hybrids vs s1 (maternal genome). But I want to remove a potential bias from reads from s2 chromosome being harder to align to s1. (The hybrid genome is composed of both s1 and s2.) I would like to correct my DE test by removing potential effect s2 vs s1. (I aligned s2 to s1 and got the count.) But my issue is, I fail to have a full rank design. I do not have any strong feelings about the library (deseq2, edgeR...).

Following the tutorial

below is the condition matrix, as you can see if I want to make the contrast ~ s2 + condition. The matrix is not full rank. I believe I could write my design as ~s2 + h - s1 ?? But I don't know how to use deseq2 for this approach.

Any approach/tips to achieve my goal here? Is it event possible? what

sample condition s2
1 s1 0
2 s1 0
3 s1 0
4 s2 1
5 s2 1
6 s2 1
7 h 0
8 h 0
9 h 0

any help is warmly appreciated!


1 Answer 1


I think you are over thinking the problem. You need to classify the genes of interest and to do that a classifier such as eggNOG is used. This will remove any bias. I understand eggNOG because it's based on protein phylogenetics and phylogenetics is what I do and I can assure you its very solid and the algorithm nicely constructed.

DESeq2 (I ain't a huge fan BTW, but avoid controversy) will happily accept eggNOG classification, here

Then its simply three tests

  1. S1 vs S2
  2. S1 vs hybrid
  3. S2 vs hybrid

I personally would perform unsupervised learning, particularly via Scanpy first on S1, S2 and hybrid separately. This will give a rough indication of the key difference of expression profiles between S1 and S2 and see where the hydrid occurs.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.