Negative interval length in GFF file

[elizabeth000]

I have been given some GFF files to process, and one of these is throwing an error when I attempt to run it through HTSeq-count:

Error occured when processing GFF file (line 26 of file ./features/promoter.gff):
Cannot subset to zero-length interval.
[Exception type: IndexError, raised in _HTSeq.pyx:348]

Indeed, on line 26, the start value is higher than the end value for the interval.

C20524	flankbed	promoter	1101	1100	.	-	.	ID=CGI_10000033;

Has anyone encountered this before?
Should I trust the quality of this GFF file or is something terribly wrong?
How would you deal with this problem?

Thanks in advance...

[Zaag]

See if it occurs more often and if it is always on the same strand, if so just turn them around.

[diego diaz]

Be careful with that GFF, it is weird to have a promoter region with a 1 base long (1100 - 1101)

[elizabeth000]

There are many such lines, with very short intervals! Yes, I'm very worried.
Unfortunately I do not know how to make gff files myself. I obtained them from the public resource at this address, where it is explained how they were made:

http://nbviewer.ipython.org/github/sr320/ipython_nb/blob/master/TJGR_OysterGenome_IGV.ipynb

Since I am completely unfamiliar with Python, I do not understand how the files were generated.

[diego diaz]

but what do you have to do? maybe there is an easier way to do it, instead using IGV

Do you have to analyze an unannotated genome?

[elizabeth000]

The integrated genomics viewer will not be sufficient for me; I have to perform calculations. The genome has been annotated, and many resources are already available.
If you think there is something strange in the gff file, perhaps there is a way to clean it up? I have very little experience with sequence data, so I need some experts' advice.

[diego diaz]

If the genome is annotated then it is probably that it is available in UCSC or Ensembl. In these repositories you could get a lot of curated genomic features, such us, repeat regions, TSSs, coding regions, intron, exons, promoters, etc, in many formats.

Biomart Ensembl website
http://www.ensembl.org/biomart/martv...030c1fd8678b01

UCSC tables

https://genome.ucsc.edu/cgi-bin/hgTables

[GenoMax]

There is a big different between predicted annotations and ones that are vetted by a human (which is why finishing a genome is so expensive).

You may want to contact the lab that generated the files (http://faculty.washington.edu/sr320/?page_id=551) and ask them about this discrepancy.

[elizabeth000]

Aha! I found the most recent versions on Ensembl. Thank you for pointing me in the right direction.