Cufflinks/cuffdiff: is this a bug?

[PFS]

I was surprised to see that in the cuffdiff output "gene_exp.diff" there are more than one row for several genes:

For example, in the results below the gene "Znhit6" has two estimated different from each other.

I thought that the file with the gene-level summary will have one row per gene, and the isoform_exp.diff can possibly have multiple row (i.e. isoforms) for the same gene.

Is this a bug or am I misunderstanding the output content?

THanks!


ENSRNOG00000030049 Znhit6 chr2:243746295-243778316 q1 q2 OK 54.7463 27.6762 -0.682138 2.92469 0.00344801 yes

ENSRNOG00000030049 Znhit6 chr2:243746295-243778316 q1 q2 OK 56.709 27.6762 -0.717361 3.09374 0.00197652 yes

ENSRNOG00000009639 Zrsr1 chr14:103571471-103667692 q1 q2 OK 37.0585 18.7667 -0.680414 3.76958 0.000163525 yes

ENSRNOG00000009639 Zrsr1 chr14:103571471-103667692 q1 q2 OK 40.0846 18.7523 -0.759679 4.16329 3.13697e-05 yes

[jasonwood]

I haven't seen this behavior before in my files, so I don't have an answer for your original question. However, I notice that your data shows the same problem as mine--the ln(fold change) column is incorrect. As i understand it, this is just supposed to be the log2 of one score divided by the other score. But it's clearly not. For instance, in your first row the two FPKM scores are almost exactly a 2:1 ratio. but the ln(fold change), instead of being -1, is only -0.68.
Does anyone have an explanation for this?
EDIT: Duh, never mind, I mixed up log2 and ln. Sorry.

[Mikie]

Hello,
I have the same issue as you and I am also interested if anyone has a solution or an explication ....
I have another question, I found this same question on the forum but no answer: Why FPKM cufflinks are different from those of cuffdiff?

Thanks!

[PFS]

It looks like this is a known "bug" occurring when transcripts do not overlap.