[BioC] edgeR cpm filtering

[Steve Lianoglou]

Hi John,

On Tue, Feb 12, 2013 at 12:45 AM, John Sperry <johnsperry51 at yahoo.com> wrote:
> Hi Jim,
>
> I posted my first question as a guest and just became a
> member today, so it’s a bit of learning curve for me on how to use the list. Sorry
> for the email and I hope I am using the list correctly now by emailing this
> address.
>
> About my edgeR question:
>
> Thank you for your elaboration and also the example. I can
> see that in your example, you have samples with 0 reads and logFC is okay, and
> that’s what logically should be. However, in my dataset, I see many cases of
> logFC of ~144269489 (or negative of ~ this value). When I check the genes, I see
> that these are the cases where all the replicates of one samples have 0 reads
> mapped to them, whereas the other groups of samples have many reads. These are
> the cases that cpm didn’t filter them. That’s why I tried to use more restrictive
> cpm filtering to get rid of these genes.
>
> Any thoughts on why this non-interpretive logFC cases happen
> are greatly appreciated.

>From the looks of your previous (original) email, it looks like you've
landed w/ a (very) strange bioconductor setup given the version of R
you are using -- you are using R-2.15.0, with edgeR version 2.6.0

Please update to the latest version of R (2.15.2 -- cuz, hey it's a
good idea), but more importantly, the latest version of edgeR:

http://bioconductor.org/packages/2.11/bioc/html/edgeR.html

You should be installing your bioconductor packages w/ the
BiocInstaller package/biocLite functionality. Unwinding yourself from
the bizarre package versions you have running to ensure that you are
on the latest bioc release train (2.11) might get a bit tricky, and
sometimes the easiest way is to blow out your R install and start from
scratch.

After you do that, do you still get such extreme logFC's?

-steve

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact