Identification of major isoform per gene from RNA-Seq

Question

[Leszek]

I would like to find out whether there are any changes in major isoform (most expressed) per gene in my time course experiment. I can get this information by parsing FPKM values from cufflinks.

Do you know of any more sophisticated tool/method capable of identifying major isoform changes between conditions?

EDIT: I've ended up writing my own script. Details in here.

EDIT2: fpkm2major.py can be downloaded from my github

[Bionic]

Could you please share your scripts!

[andrew.j.skelton73]

Just to add on to this, you can try Kallisto -> Sleuth now to analyse at the transcript level, which was unavailable when this post started.

Answer 1

[andrew.j.skelton73]

I'd use CummerBund to extract your CuffDiff results into R, and get your gene FPKM and isoform FPKM tables. From that you'll need to map Gene ID's to Transcript ID's and filter each to find your dominant isoform. Beware that due to the nature of isoform switching, you may find that different isoforms are dominent in different conditions. I don't really see what you'll achieve by looking at only the "major isoforms".

[Leszek]

Isoform switching is exactly what I would like to analyse. If there is no tool for that, I'll write something myself. Thanks!

[andrew.j.skelton73]

Ah, I've done some work on that, but it's tricky to say the least. I used EBSeq with fractional counts, I'd suggest that'd be a good stating point. You can get fractional counts out of Tuxedo, using cummerbund, the function is hard to find - check out the repCountMatrix() function.

[Leszek]

Thanks for advice. I've ended up writing my own script (fpkm2major.py) available also in my github