Cufflinks/cuffdiff and overlapping genes

[kswithers]

Hey,

I have been using cuffdiff to determine DE genes in S. pombe which has a large number of overlapping gene. I am noticing that all the genes that overlap seem to be merged as one gene in the final testing. Is this a known issue with cuffdiff and is there a way to handle it.

I'd like to keep using this software package because all the other published DE studies in S. pombe out there seem to use this pipeline. Is this something that's generally ignored in those papers or is there a simple fix i can't seem to figure out?

Thanks for you help!

[dpryan]

Do you have a stranded dataset? If not, then that's probably what's causing the gene merging. The alternative is to just use cufflinks/cuffquant for quantification with a GTF file and not try to modify existing or look for new features.

[kswithers]

It is stranded data do you think it would make difference if i use a gff3 vs gtf file?

[dpryan]

It shouldn't matter whether you use a gff3 or gtf file. Make sure you tell cufflinks and cuffdiff about your stranded library, not doing so can substantially degrade the results.

[kswithers]

yeah it was telling it the correct library type that fixed it thanks!

[kswithers]

actually that helps with cufflinks. But then cuffdiff moves back to merging them even though i tell it the library type to use. It seems to be just the the genes that overlap on the exact same value like...

SPAC19D5.06c.1 5220220 5221641
SPAC19D5.11c.1 5220220 5221383

any thoughts?

[SrCardgage]

Please excuse my ignorance. What is a "stranded dataset"?

Thank you

[kswithers]

we use the scriptseq library preparation kit so we have information on which strand the reads come from.

[dpryan]

@kswithers: Did you use the right --library-type with cuffdiff as well?

[kswithers]

yeah same one --library-type fr-secondstand

[dpryan]

No clue then. Have a look in the merged GTF and see if cuffmerge merged them.

[kswithers]

yeah something is definitely being lost in the cuffmerge part. The individual transcript.gtf files are fine then after cuffmerge this gene merging phenomena seems to be happening both when i include a reference gtf file and without it.

[bastianwur]

If I remember that right: bacteria -> polycystronic mRNA.
It's just assumed that overlapping (or very close) genes are on one mRNA, therefore they're treated as one locus.

If you take a look at your GTF file, you'll see in the last column all the identifiers, not only the "transcript locus", but also the old IDs (oID, I think).
You can tell cuffdiff in one of the options (don't ask me, I'm not at my work computer) to use these alternative IDs to do the DE.

[kswithers]

That would be awesome and an easy fix! Any idea what that option may be i can't seem to find anything in the man or manual pages

[bastianwur]

I've just checked now, and I can't find it o_O.
Maybe I did some conversions with another program, but then also no clue what I did at that time point (the merged orfs are fine for me, so I didn't bother).

I'd try to simple replace the oID entry with geneID, and the other way round. Maybe that's already enough.