[BioC] error in DEXSeq fitDispersionFunction()

[Wolfgang Huber]

Dear Christopher

better not to take such discussions off the list. 

The best thing to do is a use a current R-devel (candidate for R 3.1) from http://cran.r-project.org (see "Sources of R alpha and beta releases”), with that biocLite will automagically also go for the development versions of the Bioconductor packages.

I am not sure if this is best practice (and Martin or other Seattelites might smack my fingers), another option is
   biocLite("DEXSeq", siteRepos="http://bioconductor.org/packages/devel/bioc")
on your existing R 3.0.x, but you need to learn to know what you’re doing here, and you might also have to similarly update other packages on which DEXSeq depends.

	Wolfgang

On 10 Feb 2014, at 16:57, Odhams, Christopher <chris.odhams at kcl.ac.uk> wrote:

> Hi Wolfgang,
> 
> Thank you very much for getting back to me.
> 
> Can I ask how do I upgrade the DEXseq package?
> When I uninstall and reinstall DEXseq through RStudio, I still am on DEXSeq_1.8.0.
> e.g: 
> source("http://bioconductor.org/biocLite.R")
> biocLite("DEXSeq")
> library(DEXSeq)
> 
> I think it points to the old link.
> 
> Do I have to download and install through the package source?
> 
> Thank you,
> 
> Chris 
> 
> 
> ________________________________________
> From: Wolfgang Huber <whuber at embl.de>
> Sent: 10 February 2014 15:32
> To: Odhams, Christopher
> Cc: bioconductor at r-project.org
> Subject: Re: [BioC]  error in DEXSeq fitDispersionFunction()
> 
> Dear Chris
> 
> Thank you for the follow-up! The most recent version of DEXSeq is here: http://www.bioconductor.org/packages/devel/bioc/html/DEXSeq.html right now it is 1.9.5. so best if you could test with that. I’m afraid the .csv you wanted to attach has not made it to the list (see http://www.bioconductor.org/help/mailing-list/posting-guide the section on attachments). In addition, please also provide the code that produces the ‘issue’.
> 
>        Best wishes
>        Wolfgang
> 
> On 10 Feb 2014, at 14:47, Odhams, Christopher <chris.odhams at kcl.ac.uk> wrote:
> 
>> Hi there,
>> 
>> Has the original issue now been fixed?
>> 
>> e.g:
>> 
>> exon.counts <- fitDispersionFunction(exon.counts)
>> 
>> Error in if (sum(log(coefs/oldcoefs)^2) < 0.005) break :
>> missing value where TRUE/FALSE needed
>> In addition: Warning messages:
>> 1: In glmgam.fit(mm, disps[good], start = coefs) :
>> Too much damping - convergence tolerance not achievable
>> 2: In log(coefs/oldcoefs) : NaNs produced
>> 
>> I'm using R 3.0.1
>> x86_64-pc-linux-gnu (64-bit)
>> DEXSEQ_1.8.0
>> 
>> I've attached the csv of my test and controls.
>> 
>> The dispersion function works when I change the control / test samples to 50/50%. Now I have ~25 tests and ~150 controls and I see the above error.
>> 
>> Thanks in advance.
>> Chris?
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor