multiple lines per single gene in gene_exp.diff output of cuffdiff

[reut]

question that was posted on another forum, maybe someone here will have an answer:

after running cuffdiff, in the gene_exp.diff file there are multiple lines for the same gene, with different FPKM values.
any idea for the cause or how should I handle this to receive a single value per gene?

http://seqanswers.com/forums/showthread.php?p=34650#post34650

thanks
Reut

[sdarko]

Reut,

Have you gone back to your something.combined.gtf file to look up the entry by XLOC value?

I assumed that this was a bug and wrote a little program that compares the gene_exp.diff file and the combined.gtf and fixes the locus field in gene_exp.diff to what it should be based on the gtf file.

Has anyone else had any experience with this?

Sam

[rahilsethi]

Multiple entries of a gene/locus at Cuffdiff Differential Gene Expression Output

Hi,

I'm also getting this probelm (gene_exp.diff):

XLOC_021564 Hnf4a 3:165976692-166162683
XLOC_022796 Hnf4a 3:165976692-166162683
XLOC_022797 Hnf4a 3:165976692-166162683

When I looked at isoform_exp.diff file for Hnf4a all the mentioned isoforms like in the same region 3:165976692-166162683, but seem to be distributed among the above locus ids such that no locus id above share any isoform:
test_id gene_id gene locus
TCONS_00057332 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057337 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057338 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057339 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057340 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057341 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00057342 XLOC_021564 Hnf4a 3:165976692-166162683
TCONS_00060713 XLOC_022796 Hnf4a 3:165976692-166162683
TCONS_00060714 XLOC_022797 Hnf4a 3:165976692-166162683
TCONS_00060715 XLOC_022797 Hnf4a 3:165976692-166162683

My pipeline:

Gsnap (Reference guided Mapping, default settings) --> Remove softclip bases (script shared by rna-star author Alexander Dobin) --> Cufflinks (Reference guided; default settings) --> Cuffmerge (all samples cufflinks transcript.gtf + reference.gtf) --> Cuffdiff (default settings)

The reason to remove soft-clip bases prior to running cufflinks is to have compatibility of Gsnap mapping output with Cufflinks because cufflinks uses soft-clipped bases to do assembly leading to assembled transcripts sometimes beyond reference length and throws an error at Cuffmerge. For more info on this

https://groups.google.com/forum/#!topic/rna-star/Ta1Z2u4bPfc

Right now I am re-running the above pipeline with TopHat but I don't think it will solve this problem, because the problem is at cuffmerge, I think, in deciding to include some transcript/isoform assembly under one locus id and some in the other, within the same locus region, leading to redundancy in locus region and gene enteries.

This is not for 1 gene above. Overall in gene_exp.diff file:

Out of 25667 genes 3776 are repeated
Out of 23862 loci positions (format: chr:start-end) 5759 are repeated.

That's a significant number of genes have multiple entries.

Did anyone else have this problem as well?

Cufflinks version I used: cufflinks/2.2.1
Genome: Rat genome rn5 (Ensembl db)
Reference gtf of genes: Ensembl database version 77 for rn5


Thanks,
Rahil