Hi

Using DEXSeq I am trying to identify tissue-specific exons, and in particular retina specific exons in humans. For this purpose I run a DEXSeq analysis using 20 human retina and brain short read RNA-seq data. The Brain samples here serve as a control. So I know that DEXSeq is in particular used for detecting differential usage of exons from RNA-seq data. I think that DEXSeq could also be used to identify tissue-specific exons and in the following paper they do something similar: paper

As a first step I try do identify differentially used exons in the retina by choosing the exons with p-adj value ≤ 0.01 and by further filtering the exons with a log2 fold change ≥ 2. This results in : 8970 differentially expressed exons (using an FDR of 1%) and 4144 affected genes. In order to identify tissue specific exons I would assume that I should even filter further, by "focusing on exons that exist in one tissue but not in the other, so by using exons that have reads in one tissue (retina) and no reads at all in the other tissue (brain)." --> This way I was hooping to obtain a set of exons in each tissue that are differentially used and present in one tissue but not the other.

However from my understanding DEXSeq removes exons with 0 counts across all samples, so I am not sure how I would filter on this to obtain tissue-specific exons. Am I missing something or is this still possible? Is there maybe another paramater I should be filtering on or my strategy maybe not correct?

All tips are welcome!