Comparing different transcripts expression level of a list of genes from one condition

Question

[mr.monahton]

I have 20 RNA-Seq samples obtained from normal cells (without any treatment). I want to identify from the transcript quantification data whether for a certain gene "A" which of the isoforms (A1, A2, A3, A4, A5) show differential expression or whether there is a differential transcript usage. The idea behind this comparison is to identify whether a regulatory element present in some isoforms of this gene (and not in other isoforms of the same gene) can affect the expression or usage of these transcripts. In other words, I'm interested in finding out whether there is a significant difference in the expression of specific transcripts of the same gene in one condition. I run "salmon" on the 20 fastq files to obtain the TPM of the transcripts (in my case there are ~1400 transcripts in ~500 genes). However, I don't know how to perform this kind of comparisons or if there any R package that I could use to manage and compare TPM among transcripts of the same gene. Any help will be appreciated and thanks in advance

Answer 1

[andrew.j.skelton73]

This F1000 Article shows how to perform DTU analysis using Salmon quantification - Swimming downstream: statistical analysis of differential transcript usage following Salmon quantification

This makes use of classic packages such as DEXSeq, but optimised to use alignment-free quantifications.

Your hypothesis is pretty unclear, so I'm not sure what you're testing the different usage between. This article has the framework to make use of your input data, and perform the analysis you're interested in.

[mr.monahton]

Thank you Andrew for your suggestion. I will certainly have a look at it. The problem is that the classic packages always compare the expression between 2 conditions. In my case there is only 1 condition, and I want to compare the expression of transcripts for a certain gene to identify which transcript show the higher expression than other transcripts in a single gene. It is not easy to compare this just by looking at TPM so I was wandering if there was a tool that can handle this and how to use it

[andrew.j.skelton73]

DEXseq is most likely what you want. For a single gene, you can look at the sample-wise usage metrics, such as this visualisation (third figure down). They've analysed this with a control and knockout, but you can easily just provide an intercept of 1 design (~1) where required.