Distinguish the grafts in bulk RNA-seq

Question

[ziyan]

I have bulk RNA-seq data for a tissue and tissue-graft mixture. Is there a way to eliminate the tissue data from the tissue-graft mixture and only get the graft sequencing data?

[seidel]

Can you be more clear or specific about what you're asking? You have fastq sequence reads from libraries prepared from tissue, and libraries prepared from a mixture of tissues (i.e. does tissue-graft mean tissue plus other tissue grafted from some other region)? If one of your data sets involves a mixture of tissues, there is no effective way to separate them post library construction. By analogy, if you add a teaspoon of sewage to a barrel of wine, is it still a barrel of wine? Is there any way to separate it afterwards?

Answer 1

[Friederike]

There are some ways to approximate this, but as seidel pointed out, there's no way to "clean" the data 100% effectively unless the reads happen to be tagged for either condition.

Off the top of my head, there are two approaches you could investigate further:

1. if tissue and graft have different genetic signatures, e.g. come from different persons or even different organisms, then you can distinguish those reads that fall into regions covering loci that differ genetically
2. since you have a sample that represents the tissue alone, you could try methods that attempt deconvolution of bulk RNA-seq signal (just google that, you'll find plenty of packages)