RNAseq differential expression analysis : learning design matrix

Question

[guillaume.rbt]

Hi all,

I'm a bioinformatician and I often have to deal with RNAseq differential gene expression analysis projects.

I think I understand well the whole process to get from the raw data to the the normalized read counts but, unfortunately, due to my little statistical background, I'm having trouble dealing with the last step of differential expression.

When it comes to simple pairwise comparison between two conditions I understand the process, but when there is more complex comparisons (timelapses, multiple comparisons, including confounding effect ... ) I'm struggling for chosing the relevant design matrices.

I'm curious if anyone would know good tutorials, online courses, books, or any ressources that would allow me to learn how to get better at that.

Thank for your help,

[Benn]

Try reading the limma and edgeR user's guides. They were written for not statistician audience.

[Asaf]

I was also searching for such a guide. I ended up reading a lot of Q&A in DESeq2 support forum, they're usually well explained and written by Mike, DESeq2 developer.

[guillaume.rbt]

Yes that is also what I usually do when I'm stuck, it is really helpful !

Answer 1

[i.sudbery]

I've not actually done it myself, but people say good things about Rafael Irizarry's courses. Particularly see weeks 3 and 4 of the Introduction to Linear Models and Matrix Algebra course here and chapter 5 of his book.

[guillaume.rbt]

Thank for the tip, I will look into it :)

[noorpratap.singh]

I have gone through these courses and these actually provide a solid intuition behind the same.