cuffmerge merges multiple loci in RABT

[secda1]

Hey guys,

I am having a few problems usings cufflinks and cuffmerge in RABT mode. Everything works well, I get my merged.gtf file, but cuffmerge appears to merge some loci which are really close to each other(despite providing a ref gff file) ...
Do you have any solution to prevent that ?
Cheers


David

[Brown_lineage]

I am having the same problem.

I work on a single celled eukaryote with rather poor existing gene models, which I would like to improve. Intergenic regions are not that large and the average intron is only 142bp.

Cufflinks with the reference gtf detects some novel transcripts but merges a lot of transcripts that are not spanned by reads. I reduced max. intron to 400bp and this helped but not enough. What is weird is that the junctions.bed file shows the correct intron-exon boundaries. Can I force cufflinks to use only the junctions.bed file? In attach you see the IGV output of the mappings, transcripts.gtf of one file, the cuffmerge transcripts.gtf and the junctions.bed.

Should I switch to another package or can I make things happen with Cufflinks? Thanks for any help

[Brown_lineage]

Bump. Nobody else has had a similar problem?

[Brown_lineage]

Apparently using cuffcompare solves most of my problems. Gene models that overlap in the UTR but have different orientations are still problematic but everything else looks much better.