[colindaven]

I also tried Hmmsplicer and it seems reasonable and well documented.
We haven't got as far as checking and verifying results yet though.

[cedance]

I tried to run it. Doesn't it require a training set? And also that reads all of the same length...? It would be nice to know how good you find it.

[Simon Anders]

Simon, thanks for the clarification. Could you briefly explain (or point to a link) regarding the reason/usage of <negative binomial distribution> for variance estimation over two biological replicates?

See my post #5 in this thread.

Also, why wouldn't this package be useful, say, if I want to compare differential exon expression between two species?

At the moment, DEXSeq does not offer a good way to account for, e.g., the effect of exon length differences or the effect of sequence differences on exon coverage, and failing to take care of that might give a lot of false positives. On the other hand, Blekhman et al. did not seem to worry about this either; presumably, they considered the differences between human, chimp and macaque to be small enough to be neglected.

[ucpete]

Seconding HMMSplicer especially if you're interested in novel splice variants (it doesn't depend on potentially erroneous gene models).

software

paper

[blue mood]

I am new in RNAseq. I have the annotation and alignment of RNAseq reads by TopHat. Now I want to learn genes with alternative splicing events in each sample, e.g., the AS types and AS junctions. Is there any software that can deal with it? I am not familiar with Cufflinks.
Thanks.

[muzz56]

If you're still looking into this, you'd probably want to try MISO http://genes.mit.edu/burgelab/miso/docs/

[blue mood]

Hi,
I have read the of MISO, it seems that it requires an annotation of the alternative splicing events or isoforms. But I have no such information, I only have a predicted annotation. Am I right?

[muzz56]

Yes you need an indexed alternative events annotation file for MISO runs. For most annotated genomes, this will be in your gff/gtf files, which you can then index with the scripts provided in MISO. So to the extent that you used Tophat, you must be having an annotated genome, probably not alternative splicing annotation but something to work with. In that case I'd use de novo sequence assembly, then map back the assembled sequences to your genome of interest and use something like IGV to visualize evidence for alternative events. There are several ways to do this,and probably someone will give you a better strategy.

[enomis]

Hi there,

this thread is an old one, I know ... I just wanted to ask about Simon's statement regarding DEXSeq:

At the moment, DEXSeq does not offer a good way to account for, e.g., the effect of exon length differences or the effect of sequence differences on exon coverage, and failing to take care of that might give a lot of false positives.

Is this still true? Or does it only refer to older versions of DEXSeq?

[Simon Anders]

I write this in the context of comparing between two different species. If you all your samples are from the same species, this is not an issue.

If you compare across species, and an exon changes in length or sequence from one species to the other, you cannot test easily for differential usage. This is not a limitation of DEXSeq, but a general problem: If the exon is different, what baseline ratio of counts corresponds to "equal usage"? I don't know of any good solution for this. All papers comparing exon usage or splicing that I am aware of circumvent the issue by restricting the analysis to exons which are very well conserved.

[enomis]

I write this in the context of comparing between two different species. If you all your samples are from the same species, this is not an issue.

Ah, yes, it is stated clearly in your original posting ... I read so many threads about differential splicing analysis methods right here the last days that I mixed it up. Sorry for that.

If you compare across species, and an exon changes in length or sequence from one species to the other, you cannot test easily for differential usage. This is not a limitation of DEXSeq, but a general problem: If the exon is different, what baseline ratio of counts corresponds to "equal usage"? I don't know of any good solution for this. All papers comparing exon usage or splicing that I am aware of circumvent the issue by restricting the analysis to exons which are very well conserved.

I understand. Thank you for the clarification and your quick reply, your postings here are really useful. I'll have a closer look now on DEXSeq and the paper of Eduardo Eyras comparing different methods (http://arxiv.org/pdf/1304.5952v1.pdf) and hopefully I'll figure out a good way to analyze my data.

[gourja]

DEXSeq

DEXSeq provides differential exon usage not differential isoform usage ..If one is interested in differential isoform usage then what should be use?

[jcaswell]

I would recommend EBSeq for differential isoform usage:

http://www.biostat.wisc.edu/~kendzior/EBSEQ/

[paulorapazote]

Astalavista - too many intron retention events

Hi
I’m running Astalavista and apparently the number of intron retention events reported is not accurate, in some cases the same event is counted more than once.
Astalavista predicts alternative splicing events comparing combinations of two transcripts, example:

Transcript 1 sequence - aaacccctttiiiiiiiiiiiiiiiiiiiitttaaaccc [iii – intron retained]
Transcript 2 sequence - aaaccccttt----------------tttaaaccc [---- – intron spliced]
Transcript 3 sequence - aaaccccttt-------------iiiiitttaaaccc [---- – intron spliced; iii part retained]

What happens is Transcript 1 is reported twice as retained, two events are generated - Transcript1 xTranscript2 and Transcript1xTrancript3, but in fact there is only one intron retention event.
Any idea about this issue??
Cheers,
P

[wanfahmi]

Hi
I’m running Astalavista and apparently the number of intron retention events reported is not accurate, in some cases the same event is counted more than once.
Astalavista predicts alternative splicing events comparing combinations of two transcripts, example:

Transcript 1 sequence - aaacccctttiiiiiiiiiiiiiiiiiiiitttaaaccc [iii – intron retained]
Transcript 2 sequence - aaaccccttt----------------tttaaaccc [---- – intron spliced]
Transcript 3 sequence - aaaccccttt-------------iiiiitttaaaccc [---- – intron spliced; iii part retained]

What happens is Transcript 1 is reported twice as retained, two events are generated - Transcript1 xTranscript2 and Transcript1xTrancript3, but in fact there is only one intron retention event.
Any idea about this issue??
Cheers,
P

How do you get the output from AStalavista? I am using the commandline provide in the help section but it doesn't help. The command I used was

./astalavista -t asta -i transcript.gtf -o output.txt 

But the output was in binary file. How do I convert into file? I used astalavista-3.2 version.