Hi Dr. Linney,
The values for FPKM are relative to each experiment: the set of
transcripts/genes assembled and evaluated, the normalization methods
used. For example, when different samples are input, a different set of
isoforms per gene can result from the assembly.
The manual, FAQ, and 'how it works' help to explain in more detail, but
in short, this is expected with many of the ways that these tools are
used. Any assembly or normalization method, that impacts the FPKM
calculation, is a factor. http://cufflinks.cbcb.umd.edu
Running all the samples together is one option, to have values on the
same relative scale. Or, if you wanted to skip discovery, and you have a
reference annotation file to use as the "truth" assembly (best), this
simplified protocol might be a fit for you:
Hopefully this helps! Others are welcome to post comments/more suggestions,
On 2/15/14 8:26 AM, Elwood Linney wrote:
I am sure this has come up before and maybe I missed the answer.
If in using cuffdiff I run a sample 1(3 repeats) vs a sample 2 and
then run the same sample 1(3 repeats) against a sample 3 or a sample
4, I get different value for fpkm from sample 1 each time.
Is there something going wrong here or is there something in the
program that causes this difference?
I have seen this with online Galaxy and with the instance I have on my
The Galaxy User list should be used for the discussion of
Galaxy analysis and other features on the public server
. Please keep all replies on the list by
using "reply all" in your mail client. For discussion of
local Galaxy instances and the Galaxy source code, please
use the Galaxy Development list:
To manage your subscriptions to this and other Galaxy lists,
please use the interface at:
To search Galaxy mailing lists use the unified search at: