[Genome] GTF custom track problem ..

[Benjamin Berman]

Related to this, there's another problem i've had with the GTF  
parser.  When I have a file containing one good line and one bad line  
(as shown below) , the parser accepts it but throws the internal  
representation into some really bad state.  I can't use any of the  
pages on the browser, because I get an error:

"Error line 2 of ../trash/ct/ct_genome_5aa2_a6f8a0.bed: BED blocks  
must span chromStart to chromEnd. (chromStart + chromStarts[last] +  
blockSizes[last]) must equal chromEnd."

So I can't use the browser, "Manage Custom Tracks", or any other page  
on the system.  The only way i can get rid of the errant track is to  
hit "click here to Reset all preferences" from the main page, which  
also gets rid of all my other track settings.  Maybe from that  
internal BED file you can tell what's going on.

THanks,
ben.


On Jan 26, 2007, at 1:06 PM, Benjamin Berman wrote:
> Hi there,
>
> I'm having another problem with custom tracks, I don't know if it's  
> a bug or just confusing documentation.
>
> When I put the following GTF file into the custom tracks page, it  
> works great:
>
> chr8    hg18_ct_UserTrack       exon    121000000        
> 121010000       1000.000000     +       .        
> gene_id	"gene_idREGION1"; transcript_id "gene_idREGION1";
>
> But when I remove the "transcript_id" attribute, it chokes:
>
> chr8    hg18_ct_UserTrack       exon    121000000        
> 121010000       1000.000000     +       .       gene_id  
> "gene_idREGION1";
>
>
> Unfortunately, it doesn't choke gracefully.  It doesn't say that  
> there's an error parsing the file, or anything like that.  It goes  
> back to the "Manage Custom Tracks" page, and it looks like it was  
> successfully read in, but the track line has no "Items" or "Pos"  
> columns as it normally does (I'm attaching a screenshot).  The  
> track actually doesn't register at all; for instance, if you go to  
> the table manager it's simply not there.
>
> I went to the GTF2 specification page, it looks like maybe both the  
> "gene_id" and "transcript_id" attributes are mandatory for valid  
> GTF.  I'm not totally sure.  But your formats page says that they  
> are just "examples" of attributes one can use, it doesn't say  
> anything about them being mandatory.
>
> Anyway, i hope you could change the parser so it at least complains  
> correctly if you enter a GTF without the correct format.  This  
> should save people frustration :)  You could also change your  
> documentation page if it doesn't accurately reflect the GTF  
> requirements.
>
> Thanks!
>
> ben.
>
>
> <CustomTracks-GTF-problem.jpg>
>
>
> -----
> Ben Berman, PhD
> Postdoctoral Fellow, Preventive Medicine
> Keck School of Medicine of USC
> 1441 Eastlake Ave., Rm NOR 4423
> Los Angeles, CA 90033
>
>