cuffdiff dispersion cutting off low values

[jake13]

I've been running cuffdiff on some paired end rna seq samples with biological replicates. Before I started analyzing them I checked the dispersion plots using cummeRbund. For some reason it is forming a sharp cutoff which looks different from the examples on the websites. I have tried mapping both with tophat2.0.9 and STAR aligner and got identical results.

I have been using cuffdiff with the default options except for -b and using the genes.gtf file provided by the Illumina iGenomes download. Does anyone have ideas on what might be going wrong?
Thanks

[jake13]

An update for anyone else running into this problem. It seems to be caused by the gtf file that I was using. I tried grabbing the refseq gtf file from the UCSC table browser as well as the genes.gtf fill that's included in the Illumina iGenomes package. I ran identical reads with identical settings through tophat/cuffdiff except for using these two different gtf files. The UCSC one is on the left and looks correct, the iGenomes one is on the right and has the weird cutoff. I have no idea why this is occurring. Does anyone have any thoughts?

[blouro]

Up with this topic, because I'm getting similar problem. Used the Tophat and tuxedo tools pipeline, and now looking at the dispersion plots I'm getting a strange profile in two aspects, the values are way off the expected range, and the dispersion profile seem way zoomed out due to few lower values. Please see attached plot and give me your opinion, it will be much appreciated!

P.s For the cuffdiff analysis I used the merged gtf file created in the cuffmerge step.