Hello,

I have some analysis recently with DEXSeq package, on RNA-seq data (paired-end, mouse), following the very nicely written vignette and manual that are provided (including the DEXSeq Python scripts to build the counts matrix etc).

My issue though concerns the "Gene Model" that is drawn in the plots output. Although for many genes, this gene model corresponds indeed to the RefSeq given in UCSC or Ensembl, for many others, the gene model drawn on the top does not seem to make really sense.

At the beginning, I thought that this could be due to loci where there are multiple isoforms with different exons overlapping more or less etc, and that the Gene Model would thus correspond to a kind of "sum" of all the existing exons. But even like that, the exons of the Gene Model clearly don't "fit" to the exons of all the shown isoforms and transcripts on that region.

My question: how is the Gene Model built?

Thanks,

H.

ps: since it does not seem that pictures can be inserted in the text directly, I can send some pictures by email. I have for instance a locus where it says that the Gene Model has something like 42 exons, while this is clearly not true (at least, from the RefSeq info). The exons E37, E38 and E39 of that locus (same for E26/E27/E28, or E31/E32) seem to fit only to introns or to a subsection of the last exon of a given isoform.

Dear Anonymous,

Did you have a look at the paper accompanying DEXSeq, by Anders, Reyes, Huber; Detecting differential usage of exons from RNA-Seq data. Genome Research, 22:2008-2017, 2012 and in particular the section: Method /
Preparation: Flattening gene models and counting reads.

If your problem persists, then please do send pictures of what you mean. It's easy to insert them in posts, e.g. see below. Also, the code that you ran to arrive at your counting bins.

Best, Wolfgang

Dear Wolfang,

Thanks for your help. Indeed, I had a look of course to your paper and to the flattened gene model. But I think I was just confused by the fact that in the gene model output, the sections are called/numbered by "exons", whereas they are not really exons per se. So, we are not actually looking at "Differentially Expressed Exons" per se, but rather at differential expression of "regions"  (purple highlight) that correspond to "sections" (sorry, I don't know how to name correctly the sectioning of the model gene based on the known isoforms) of exons.

Anyway, this answers my question :D

Thanks,

H.

Below an example of output. So the numbering in the gene model corresponds to all the sections from the different isoforms. Although the output says "total_exons"= 42, there are not 42 different exons found for this locus. But indeed, the comparisons of the exons starts and ends for all the isoforms and other overlapping genes make 42 different sections.

So, as expected, the pictures (.png, .jpeg, ~50kb) don't seem to appear in my posts, probably because of rights issues (?). Apologies.