GTF file for cuffdiff 0.9.1

[kalidaemon]

The GTF file I'm using to run cuffdiff has transcript IDs but no p_ids. Consequently, cuffdiff is unable to make the cds, promoters, splicing, and tss_groups files. Is there a database where I could get an improved GTF file? If not, what table schema in the UCSC genome browser have people used to construct their GTF files?

Thanks!

[Thomas Doktor]

Have you run cuffcompare on your samples first? Cuffcompare attaches p_ids and tss_ids to the combined GTF file that you can then use as input for Cuffdiff.

[kalidaemon]

Thanks! I tried using the .gtf file from cuffcompare as my reference gtf for cuffdiff as you suggested. This solved some problems, but created others. The isoforms, promoters, splicing, and tss files are now populated, but the cds files still aren't. The other thing that happened is that there were no recognizable gene names in any of the files created by cuffdiff with the cuffcompare .gtf file. Instead, the gene names were "XLOC..". I'm thinking there is a problem with my reference gtf file that I used in cuffcompare. Where can I find a better reference gtf for mm9?

[ramouz87]

Hi,
I'm facing the same issue with the mouse gtf file,
It will be good that the gtf of major organism will be made available of the cufflinks page.
Best,
Ramzi

[fkuo]

from my understanding, to have p_id you need to run cuffcompare with the -s option. Also no gene names are showing up probably because the gtf that you are supplying it does not have a gene_name attribute in the 9th column, you should try the Ensembl GTF, that one has gene names http://uswest.ensembl.org/info/data/ftp/index.html

[kalidaemon]

fkuo: Thanks! I tried running cuffcompare with the -s option and was able to generate a p_id. Unfortunately, my troubles didn't stop there. My combined.gtf file contained tss and p ids that didn't really make much sense. This resulted in lots of NO TEST error messages when I ran cuffdiff. What did you use as your -r .gtf files? other .gtf? Also, did you use the -p option? if so, how do you specify the prefix?

[fkuo]

hi kalidaemon,

for the -r, I used a combined reference gtf (UCSC, Ensembl, Refseq). For the -p option, you just used -p4 for 4 threads or --num-threads 4. hope this helps!

[labrat73]

no p_id attribute

fkuo: Thanks! I tried running cuffcompare with the -s option and was able to generate a p_id. Unfortunately, my troubles didn't stop there. My combined.gtf file contained tss and p ids that didn't really make much sense. This resulted in lots of NO TEST error messages when I ran cuffdiff. What did you use as your -r .gtf files? other .gtf? Also, did you use the -p option? if so, how do you specify the prefix?

Hi kalidaemon,

I see that you ran cuffcompare with the -s option and was able to generate a p_id. I tried this and still wasn't able to generate the attribute. Could you offer any tips?
Many thanks!