Search
Question: DESeq2 old versions
0
gravatar for riccardo
18 months ago by
riccardo30
riccardo30 wrote:

Hi,I am a PhD student and i would like to know if it is possible to find online the previuos versions of DESeq2 and where.Thank you.

Best regards.

Riccardo

ADD COMMENTlink modified 17 months ago • written 18 months ago by riccardo30
0
gravatar for Wolfgang Huber
18 months ago by
EMBL European Molecular Biology Laboratory
Wolfgang Huber13k wrote:

https://bioconductor.org/packages/3.0/bioc/html/DESeq2.html, https://bioconductor.org/packages/3.1/bioc/html/DESeq2.html, etc. (just edit the Bioconductor release number, 3.0, 3.1, etc., in the URL).

If by "old versions" you also mean DESeq, then e.g. https://bioconductor.org/packages/2.6/bioc/html/DESeq.html

Finally, you can check out previous versions from subversion, as described at the bottom of each package landing page (e.g. https://bioconductor.org/packages/devel/bioc/html/DESeq2.html), by specifying the revision number (-r flag). And something analogous also exists on the git mirror: https://github.com/Bioconductor-mirror/DESeq2 -- the "Branch" button.

 

 

 

 

ADD COMMENTlink modified 18 months ago • written 18 months ago by Wolfgang Huber13k
0
gravatar for riccardo
17 months ago by
riccardo30
riccardo30 wrote:

Thank you. But is it possible to download a specific version, for example the 1.9.16?

ADD COMMENTlink modified 17 months ago • written 17 months ago by riccardo30
0
gravatar for Martin Morgan
17 months ago by
Martin Morgan ♦♦ 20k
United States
Martin Morgan ♦♦ 20k wrote:

Bioconductor packages are maintained in 'subversion' (svn) version control. Basic instructions for accessing svn are at http://bioconductor.org/developers/how-to/source-control/. The DESeq2 package svn url is https://hedgehog.fhcrc.org/bioconductor/trunk/madman/Rpacks/DESeq2. Running

svn diff -log https://hedgehog.fhcrc.org/bioconductor/trunk/madman/Rpacks/DESeq2/DESCRIPTION > DESeq2.DESCRIPTION.diff

shows that 1.9.16 was associated with revision 105633 (this is an approximation, there are situations where the actual svn revision exposed to the user differs from this).

With the revision in hand,

svn co -r105633 https://hedgehog.fhcrc.org/bioconductor/trunk/madman/Rpacks/DESeq2

will get the snapshot of the package, and it can be installed with R CMD INSTALL or from within R using install.packages().

When I load the current release version of DESeq2, sessionInfo() tells me that 10 packages are attached to the search path, and 35 packages are loaded. To reproduce the session using 1.9.16 (I guess that is your motivation, to reproduce an earlier analysis) requires that the version of R, and all of these package versions match the version in use for the analysis that you are trying to reproduce. To do this in retrospect is very difficult, and requires as a minimum starting point the sessionInfo() or other capture of the old session. There are several approaches to doing this in a prospective fashion (e.g., packrat; docker or other images), as has been discussed elsewhere in this and other forums.

Actually, this StackOverflow question includes a suggestion using git. DESeq2 is mirrored in github at https://github.com/Bioconductor-mirror/DESeq2, so I did

git clone git@github.com:Bioconductor-mirror/DESeq2.git
cd DESeq2/
git log -p DESCRIPTION
git checkout daebada7f87824225cfffe7f1793594eaaff8f51

for a much faster approach to the revision. (This leads to a 'detached HEAD' state in the git repo, whatever that is! git checkout master when done).

ADD COMMENTlink modified 17 months ago • written 17 months ago by Martin Morgan ♦♦ 20k

Thanks Martin. Also to throw in the obvious: 1.9.16 is a version in the middle of a development cycle. For package version number x.y.z, if y is odd, this means development branch versions, which are still very much in flux before the end of the cycle and you get y+1 as the release version. In the middle of a development cycle, there very well could be bugs still being ironed out from newly introduced features. So while it's good for developers to use the development branch, for proper analyses, you'd really want to use release branch, i.e. even numbered y versions of software.

ADD REPLYlink written 17 months ago by Michael Love12k
0
gravatar for riccardo
17 months ago by
riccardo30
riccardo30 wrote:

Thank you.

Riccardo

ADD COMMENTlink written 17 months ago by riccardo30
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 270 users visited in the last hour