Cufflinks merges adjacent genes

[proteomania]

Hi all,

I am working on the RNA-Seq of an organism with very short intergenic distance (e.g. 50bp). I am using the TopHat/Cufflinks package for my analysis, and I found out Cufflinks tends to merge adjacent genes as a single trasfrag, when the intergenic distance is short. I wonder whether there's an option to adjust the cutoff for the coverage gap size for the definition of a transfrag.

Thanks in advance.

cheers,
Chung

[GKM]

There's a coverage option for filtering out transcripts with "retained introns" that are assembled due to complete coverage of the intron with reads from pre-spliced mRNAs. However, this isn't going to work in your case as the intergenic regions aren't going to be spanned by junctions and I don't think there is any option that does what you want.

Genes that are on different strands should be assembled relatively correctly, as there the orientation of the junctions will tell Cufflinks what to do. But for genes in the same orientation, you are out of luck, especially if they are at about the same expression levels.

What you could do is look at you transcript models and try to find those that have two tandem ORFs, those are likely candidates for merged genes. Also, you could try doing ChIP-Seq for histone marks associated with promoters or preinitiation complex components.