[BioC] Latest Stable Limma Version

[Gordon K Smyth]

Dear Alejandro,

To add to Herve's comments:

1. Both versions of limma are stable.  They wouldn't be on the public repositories for R 2.4.1 if
they weren't.  Naturally I recommend you use the latest version -- that's why it's there.

2. Actually install.packages("limma") and biocLite("limma") both give you the same version (2.9.8)
of limma, so there is normally no confusion.  You could only get limma 2.8.1 if you take special
action to reset the CRAN mirror to point to bioconductor instead of CRAN.

3. As Herve has said, limma is on both the CRAN and Bioconductor repositories.  This is explained
in Section 2.2 on installation in the Limma User's Guide.  The Bioconductor version has to satisfy
Bioconductor rules, which means no changes to the user interface between six-monthly releases. 
Hence stable updates to the package between BioC releases can only happen on CRAN rather than
Bioconductor.  You can be sure limma works with the current release of Bioconductor, otherwise it
won't be posted to CRAN.

4. It is potentially confusing that biocLite("limma") sometimes installs from CRAN rather than
Bioconductor, but I'm happy with this behaviour.

5. The reason for the 2.9.8 version number is that limma 2.9.8 on CRAN is identical to limma 2.9.8
on the Bioconductor developmental repository.  If at any time they are different, they'll have
different version numbers.

Best wishes
Gordon

> Date: Mon, 29 Jan 2007 12:51:08 -0800
> From: Herve Pages <hpages at fhcrc.org>
> Subject: Re: [BioC] Latest Stable Limma Version
> To: Alejandro <ale.sola at integromics.com>
> Cc: bioconductor at stat.math.ethz.ch
>
> Hi Alejandro,
>
> Alejandro wrote:
>> Hello,
>> I've tried to install the latest stable version of Limma.
>> With "biocLite()" command I get Limma 2.9.8 but if I use the
>> "install.package" command I get Limma 2.8.1.
>> How can I get the latest "stable" version of Limma package?
>>
>> With the following commands I get Limma 2.9.8 (Is it a stable version?)
>>
>>     source("http://bioconductor.org/biocLite.R")
>>     biocLite(c("limma"))
>>
>> With the following command I get Limma 2.8.1
>>
>>     install.packages('limma',repos="http://www.bioconductor.org/")
>
>
> Short answer: I would say that it's probably OK to use limma 2.9.8 with R-2.4
> (even if it "looks" like a Bioc devel package).
>
> Long answer:
> You've just given us an example of why "dual" hosting of packages can sometimes
> lead to confusion. Let me explain: limma is a Bioconductor package but, at the
> same time, is hosted on CRAN. The version on CRAN is 2.9.8, but there is no 100%
> safe way to know which version of Bioconductor it is supposed to work with.
> The middle number in 2.9.8 is odd and, according to Bioconductor version standards,
> this suggests a devel (unstable) package. But, since this version is on CRAN, it's
> probably the case that the parity of the middle number is irrelevant and that the
> package is stable _independently_ of which version of R/BioC you are using.
> This is because, generally speaking, CRAN source packages are not tied to a particular
> version of R (except of course when they make use of something like "R (>= 2.3.1)"
> or "R (<= 2.4.0)" in their Depends field).
>
> So now why do biocLite() and install.packages() pick up a different version of limma?
>
> Officially, the limma package in BioC 1.9 (release) is 2.8.1. That was the latest
> available version at the time Bioconductor 1.9 was released back in October 2006.
> After a BioC release we create a "release" branch in our Subversion repository so
> people can maintain separately the "release" and the "devel" version of their package.
> The limma package in the "release" branch has not changed since then and this is the
> version you get when you do
>   install.packages('limma', repos="http://www.bioconductor.org/")
> or
>   install.packages('limma', repos="http://bioconductor.org/packages/1.9/bioc")
>
> In the case of the limma package, using biocLite is equivalent to doing
>   install.packages('limma',
>     repos=c("http://bioconductor.org/packages/1.9/bioc", "http://cran.fhcrc.org/"))
> What's happening is that install.packages sees both versions of limma (2.8.1
> in http://bioconductor.org/packages/1.9/bioc and 2.9.8 in http://cran.fhcrc.org)
> and picks up the higher version (which is probably the right thing to do).
>
> Cheers,
> H.
>
>
>>
>>
>> My R version is 2.4.1.
>>
>> Thank you very much,
>> Alejandro