CuffLinks / CuffDiff GTF Compatibility

[Bio.X2Y]

Hi,

I want to use CuffDiff to call DE between two samples that I've aligned with TopHat.

Rather than creating a GTF via CuffLinks, I used one from UCSC (RefSeq). Since the manual suggests the GTF can be from sources other than CuffLinks, I assume this isn't a problem?

One thing that confuses me, however, is that CuffLinks seems to creating 0-based coordinates in its GTF (as per manual), while the GTF 2.2 standard calls for 1-based coordinates.

Since the UCSC GTF uses 1-based, does that mean UCSC GTF files are incompatible with CuffDiff?

Thanks,
Bio

[Bio.X2Y]

I've raised this question on the Bowtie sourceforge (I can't find a place to raise CuffLinks specific questions/bugs):

https://sourceforge.net/tracker/?func=detail&aid=3017153&group_id=236897&atid=1101606

Text:
Am I correct in thinking that the CuffLinks GTF files are non-standard? They use 0-based coordinates as per the manual, while GTF2.2 requires 1-based. Does this mean that CuffDiff is not compatible with UCSC or Ensembl GTFs?

[Marcelo_Alejandro]

While describing the start column of the transcripts.gtf file, cufflinks manual (http://cufflinks.cbcb.umd.edu/manual...fflinks_output) states:

The leftmost coordinate of this record (where 1 is the leftmost possible coordinate)

Therefore, it should be 1-based.