How to remove the RIN (RNA integrity number) effect for RNA-seq data?

Question

[Wayne Lee]

Holle everyone! I am a beginner in RNA-seq data analysis. In my recently work, I face a problem: my RNA-seq data has significant correlation with sample's RIN (RNA integrity number), and I want to remove this effect. Some papers indicate that one can use a linear model to normalize this technical covariate, but I do not know how to do it. If anybody can help me about this? Or if there have any other methods can sovel this problem? Thank you very much!

Answer 1

[Devon Ryan]

Don't bother actually removing it, just put it in as a covariate in your model.

[Wayne Lee]

Thank you answer this simple questure! But, I don't quite understand what your mean. Can you tell me more details? Or can you supply me some material or papers about this? thank you!

[Devon Ryan]

You can google "DESeq2 nuisance covariate" for various examples.

[Wayne Lee]

THX, it's helpful !!

Answer 2

[Michele Busby]

You may want to try looking at the supplements for the GTEx papers. They did a lot of work on that. It also depends a lot on the protocol. If it is PolyA selected you will get biases against short genes because you will lose the 5' end. If it is RNA depleted or hybrid capture your bias will be different, so make sure the model you are looking at is applicable. The GTEx, I believe, was mostly poly-A kits.