[BioC] DEXseq: several "1" and "NA" in padjust column

Steve Lianoglou mailinglist.honeypot at gmail.com
Tue Feb 5 12:17:52 CET 2013


Hi Vincenzo,

On Tue, Feb 5, 2013 at 5:11 AM, Vincenzo Capece <vivo0304 at gmail.com> wrote:
> Dear bioconductorers,
> I am using DEXseq for the splicing analysis.
> The pipeline works perfectly, but I have some doubts about the output: on
> 351000 exons I have just 10 padjust significant ones.
> Googling I found this thread, in which the user developed my same pipeline
> and in which she has the same problem:
> https://stat.ethz.ch/pipermail/bioconductor/2012-February/043476.html
> This thread finished in a "private" way, but I need to know if there are
> any bugs(I think fixed, because It past one year);

The thread actually finished off "publicly." Perhaps you thing it went
private because you see messages like "embedded and
charset-unspecified text was scrubbed" in the emails of the thread
that follow the one you posted?

If you click on the URL below, it's there:
https://stat.ethz.ch/pipermail/bioconductor/attachments/20120213/5f13a499/attachment.pl

That having been said, it's hard to say what (if anything) is going
wrong with your analysis without you posting any source code, but you
can calculate the adjust pvalues yourself.

After running the testForDEU step, the fData(ecs)$pvalue and
fData(ecs)$padjust columns should be filled. One could calculate the
padjust column manually by doing:

adjusted <- p.adjust(fData(ecs)$pvalue, 'BH')

but if the values in the pvalue column are NA, so too will be the
adjust pvalues.

If you have many NA pvalues after your testForDEU, then there might be
several reasons why they are there:

(1) These exons have been marked as "untestable" (fData(ecs)$testable)
during the estimateDispersions step, which might be due to:
  (a) read counts for the exon below the "minCount" required per exon
  (b) the exon is in a gene with more than "maxExon" exons
  (c) the gene only has one exon?
(2) The calculation of the pvalue errored and didn't return anything
-- not sure why this would happen so frequently, though.
(3) ... ? ...

> moreover I am interested
> knowing if the problem was at "user level" or "source code level" and, in
> the first case, how she fixed the bug.

I'm not sure if this can be answered without your data, but you will
first need to post the code yo used.

As for the original poster that you mentioned -- you'll see how her
problem was fixed when you click on the URL's in the messages that
look "hidden" from view.

> Thanks in advance and congratulation for your package: it is awesome.

Hear, hear!

HTH,
-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



More information about the Bioconductor mailing list