prediction-free exon usage analysis

Question

[Assa Yeroslaviz]

Hi,

we are working on a RNAseq data set, for which we would like to do an alternative splicing analysis on the exon level.

The group of genes we are interested in are unfortunately not so well annotated, so I am looking for a tool for such analysis, which not necessarily must have a predicted transcript list like a gtf file.

I know that DEXSeq can do this kind of analysis (as well as many other tools such as GenomeGraphs (R), SplicingCompass, MATS, SpliceR, etc.)

As far as I could tell from reading the Vignette from DEXSeq and this example for GenomeGraphs on analysing genomic data for visualization of differential expression (thanks, Neil), I figure out, that most of them needs the exon information given from the beginning.

Are there any tools who don't need this information?

Is it possible to run such an analysis without the prior prediction of the exons, I mean without the predicted transcripts?

Other problem with this data set is that we don't have any replica. So I am looking for a tool who can do all this and still doesn't work only with replica

thanks,

Assa

Answer 1

[Devon Ryan]

FYI, differential exon usage is slightly different than looking at alternative splicing in that the former is more of a much easier to compute/debug proxy for the latter.

If you have a poorly annotated genome you might want to use something like cufflinks. While you could then use cuffdiff to look at differences between samples, I wouldn't bother without replicates (the results would be too unreliable with any tool).

Regarding DEXseq, there's no way to make it work without replicates and prior exon information.

[Assa Yeroslaviz]

Thanks for the answer and sorry for the poor formulated question. My genome is not poorly annotated, it is the drosophila genome, but unfortunately the group of genes we are interested in are not so well known.

Are there any tools which can work without replica? I know about the problems and the lack of reliability in doing such analyses without replica, but this is what I have got, so I need to do the best out of it :-)

[Devon Ryan]

Cuffdiff will run without replicates.

[Assa Yeroslaviz]

thanks, are there any other user friendlier tools? :-)

[Devon Ryan]

Not that I know of, but hopefully others will reply with such examples :)

If it helps, Galaxy supports cufflinks/cuffdiff.