Bug? duplicated genes in cufflinks output genes.expr

[silin284]

Hi

When i supplied a reference gtf to cufflinks (-G), i found there are duplicated geneID in the output "genes.expr". That is a bit weird to me and it is very rare (3 out of 50k genes). I checked those 3 and it turns out that cufflink consider their isoforms as individual genes but still use the same gene_id supplied in the gtf file. All these 3 genes have a common characteristics. The genome positions of each isoform's transcript/exon/CDS are completely different. I guess cufflink use this information to judge whether different transcripts belongs to the same gene instead of using the gene_id information supplied in gtf.

I can remove them by hand but is there a way to "force" cufflinks to recognize them as a single gene?

cheers
silin

original GTF file
chr06 SZ transcript 3851140 3853473 . + . gene_id "Os06g07923"; transcript_id "Os06g07923.2";
chr06 SZ CDS 3851140 3851247 . + 0 gene_id "Os06g07923"; transcript_id "Os06g07923.2";
chr06 SZ CDS 3853062 3853304 . + 0 gene_id "Os06g07923"; transcript_id "Os06g07923.2";
chr06 SZ exon 3853305 3853473 . + . gene_id "Os06g07923"; transcript_id "Os06g07923.2";
###
chr06 SZ transcript 3851392 3852964 . + . gene_id "Os06g07923"; transcript_id "Os06g07923.1";
chr06 SZ exon 3851392 3851900 . + . gene_id "Os06g07923"; transcript_id "Os06g07923.1";
chr06 SZ CDS 3851901 3852434 . + 0 gene_id "Os06g07923"; transcript_id "Os06g07923.1";
chr06 SZ exon 3852435 3852964 . + . gene_id "Os06g07923"; transcript_id "Os06g07923.1";

cufflinks output "genes.expr"
Os06g07923 141826 chr06 3851139 3853473 0 0 0 OK
Os06g07923 141826 chr06 3851391 3852964 0 0 0 OK

[apadr007]

I have the same question. Why is cufflinks repeating genes?

[kenphi]

Dear silin

I think this is because in your reference annotation there are "unrelated" transcripts annotated to the same gene. I noticed that this happens, when there are independent transcript groups, i.e. groups of transcripts that do not overlap in exon coordinates. The can be side-by-side or one in the intron of the other. Some examples are in Ensembl 64

ENSMUSG00000086255
ENSMUSG00000062352
ENSMUSG00000021879
ENSMUSG00000033705
ENSMUSG00000087461
ENSMUSG00000022105
ENSMUSG00000073791
ENSMUSG00000052675
ENSMUSG00000055407
ENSMUSG00000056856
ENSMUSG00000027203

In some of these cases, I would say that Ensembl didn't follow its own guidelines, to assign the same gene identifier to transcripts with overlapping position, because there are clearly independent clusters.

I keep them and use the gene_id column of cufflinks to make tables unique.

Philip

[emanlee]

Another thread on this issue:

http://seqanswers.com/forums/showthread.php?t=5224

A solution based on mgogol's code:

https://sourceforge.net/projects/collapsefpkm/files/?source=navbar