Identifying composition of cufflinks transfrag

[ega2d]

Hi all,

This question has been dogging me for quite some time and I'm hoping the collective wisdom of seqanswers will be able to answer my question.

Given a specific TCONS_ID (transfrag identified by Cufflinks), is there a good (i.e. fast) way to pinpoint what exactly about that transfrag is unique??? Aside from making a really pretty table of "total splice events", it seems like transcript composition would a top priority for most people doing RNAseq...

Thanks in advance!

[RockChalkJayhawk]

This is actually a pretty annoying problem because most of the transcripts I have seen are only different by a few bases, usually at the end of an exon. These arise from the inability to split short reads in the correct place and subsequently map them correctly.

One way to do it could be to use mergeBed from the BEDtools suite. You can specify to keep the names. For pieces of transcripts that are unique to only one transcriptional unit, they will only have one name attached. If there are more, then they will have each transcript name separated by a semicolon. Then use :

grep -v ";" file > unique transcript positions

[ega2d]

Hey there,

THANKS SO MUCH FOR THE REPLY. I am finding the same instances where these splicing events only different by a few nt as well. I thought that if I ran the alignment with 0 mismatches it would be better, but still I would really love to figure out which ones are real and which ones are misalignments.

I think I follow your suggestion about identifying unique exon coordinates, but please tell me if this is what you meant:

mergeBed on combined.gtf to remove spurious "alternative splicing events" that are really only part of a known exon

then re-run cuffdiff?

Thanks again for answering, like I said this problem had been driving me crazy.

[RockChalkJayhawk]

You can't run cuffdiff on that kind of file. This is only for finding the unique parts of transcripts.

You SHOULD NOT require 0 mismatches. This introduces another issue. The problem will not go away. What if your gene has a real genetic polymorphism (or an error in the reference) near the splice site? It may not get any reads mapped to this area, and you wouldn't see the (full) transcript.

[ega2d]

Yeah, I realized my mistake about the bed file after I posted. I also have ran my analyses with the default 1-2 mismatches and honestly, didn't see that much of a change.

My real question is what file to run mergeBed with?

Thanks again for your reply!