Cuffmerge output: merged.gtf and transcript.gtf always vary ?

[Chirag]

Hi there,

Could anyone please help me to understand the difference between two output files of cuffmerge, merged.gtf (which is supposed to be final output) and trasncript.gtf.

I used cufflink to reconstruct a gene model from 3 time points and finally merged them using cuffmerge (with reference annotation). Two output files of cuffmerge are trasncript.gtf and merged.gtf. among others. Isn't the trasncript.gtf and merged.gtf suppose to be similar, or why aren't they similar?

I converted both .gtf files into .bed, and overlapped them, i found around 2000 transcripts from transcript.gtf were missing in merged.gtf ? Then i looked into those missing cases (in merged.gtf), those trasncripts have nice coverage and gene model is also present in individual time point.

So, why do we see this difference ?

Thank you for your help in advance !!

regards
Chirag

[sdriscoll]

i'm not sure of the answer but the transcripts.gtf file may be an intermediate step in the process of generating merged.gtf.

All I know is the file you want to use after running cuffmerge is merged.gtf.

[smcinturf]

The Tophat paper that was published in Nature Protocols says to use the merged.gtf, but when consulting the users guide on the cufflinks website it says to use the <transcripts.gtf>. For that reason I am still suspicious of what file to use.

[Chirag]

merged.gtf is the file to be used.
transcript.gtf file can contain multiple transcripts which are redundant (ie, have similat splicing patterns), which cuffmerge merges them.

[potato84]

It would be nice if somebody can explain what transcripts.gtf is.

[Chirag]

In my opinion: Eg: If a gene has multiple isoforms (which are coming multiple samples), all these isoforms will be in the trasncripts.gtf. If any isoforms have the exact splicing signals, then cuffmerge merges them and those redundant transcripts are not anymore on merged.gtf. So, i would suggest, or i rather use, merged.gtf.

Secondly, try to filter merged.gtf based on read count, since many transcripts come up in merged.gtf (which is basically coming from reference gene annotation).

[ZhigangLi]

I have same question. Transcripts missing in the merged.gtf are not only isoforms, but also genes. I assembled my RNA-seq data de nove and blasted result against the genes annotated in the two files, respectively. And I found that genes in transcripts.gtf fils have more hits than that in merged.gtf. Therefore, I think those genes in transcripts.gtf (not in merged.gtf) are not abundant.