Cuffdiff p-value problem

[sazz]

Hello all,

I run the Tuxedo pipeline to find differentially expressed genes btw 2 phenotypes (3 replicates for each) but in the end I get p-values limited to "0.00005" in CuffDiff output, nothing smaller; 1264 of 23779 genes (human) have that limitation so this is not their actual p-value, most of them should be much more significant. I have the latest version of CuffDiff (2.1.1). In the older versions I didn't have such a problem.

I don't know what I miss but I appreciate if you can help.

[Gus]

That is crazy because I was logging on to ask this exact question.

I am doing the same experiment in four species and i do not have the problem with all species. I will attach links to the resulting volcano plots from a species with the problem.

Can anyone PLEASE offer explanations I am at a loss...

http://previews.figshare.com/1057598/preview_1057598.jpg

[Gus]

OK i got so excited that someone was asking the same thing as me that I neglected to do a search first...

Upon searching, perhaps the answer is in this thread: http://seqanswers.com/forums/showthr...fdiff+p-values

Sounds like it is due to a change in the way cuffdiff does its p-values by default (now through empirical sampling? ).

Can others confirm this?

I will need to go back and see which version I used for the species that does not show this cut off.

Gus


PS: I checked my records and this is indeed the case. The species that doe snot show the cutoff was run with 2.0.2.

[sazz]

Thanks Gus, I am a little relieved.