[DESeq] versions, conds definition, estimateDispersions vs estimateVarianceFunctions

[Simon Anders]

I found it here

http://bioconductor.org/packages/2.9...ews/DESeq/NEWS

Right. Also note the new section at the end of the vignette summerizing differences between the current DESeq version and the method described in the paper.

[marcowanger]

I found it here

http://bioconductor.org/packages/2.9/bioc/news/DESeq/NEWS

[marcowanger]

1) The biocLite script ties Bioconductor version to R versions, i.e., depending on your R version it downloads different package version; specifically, for R 2.13, it pulls from Bioc 2.8 and for R 2.14 (just released) it pulls from Bioc 2.9 (currently 'deve', will be release in a few days). Sometimes you can circumvent that by installing a package manually but this might require that you do the same with the packages it depends on.

3) No, sorry. You need to change the command in your scripts.

4) Non-factors get automatically converted to factors by 'newCountDataSet' so it does not matter.

Is there any change-log for each version of DESeq? I was working with 1.2.1 and now the release is 1.6.0.

[Simon Anders]

1) The biocLite script ties Bioconductor version to R versions, i.e., depending on your R version it downloads different package version; specifically, for R 2.13, it pulls from Bioc 2.8 and for R 2.14 (just released) it pulls from Bioc 2.9 (currently 'deve', will be release in a few days). Sometimes you can circumvent that by installing a package manually but this might require that you do the same with the packages it depends on.

3) No, sorry. You need to change the command in your scripts.

4) Non-factors get automatically converted to factors by 'newCountDataSet' so it does not matter.

[DanR]

Hi,

1)
The session info from the vignette shows Anders is running R2.14.0 not R2.13.2 which you installed. The version of DESeq are different too (Anders: 1.5.35, you are running 1.4.1).
I think you may need to upgrade the version of R and bioconductor before your default repository points to the newest version of DESeq.

2) If the functions are new, you would not find "estimateDispersions" in an old version.

4) I think the factor just informs R that the vector has variables that are factors == like "normal" and "cancer"

Dan

[Azazel]

1.) On a fresh install of R 2.13.2 (Win 7), I am trying to install the current version of DESeq from http://bioconductor.org/packages/dev...tml/DESeq.html.

According to that page, the version to be installed should be version 1.5.35.

However when I install it with these commands:

source("http://bioconductor.org/biocLite.R")
biocLite("DESeq")

and then check the installed version (installed.packages()[,c("Package","Version")]) I get:
DESeq "DESeq" "1.4.1"

2.) In the manual I noticed that there seems to be a new function, called "estimateDispersions" which seems to be similar to what was previously "estimateVarianceFunctions". I am referring to the manual PDF dated "October 29, 2011". However this new function does not seem to exist in my installation (this could be related to the above point of it being version 1.4.1):

> cds <- estimateDispersions(cds, method="blind", sharingMode="fit-only")
Error: could not find function "estimateDispersions"

3.) I have written a couple of scripts that contain calls to "estimateVarianceFunctions" - will they still work with future versions of DESeq?

4.) One thing I noted about the conds vector. Prevously (manual from "April 21, 2011") it was defined like this:

conds <- c( "T", "T", "T", "Tb", "N", "N" )

But now it's defined like this:

conds <- factor( c( "treated", "treated", "untreated", "untreated" ) )

So previously it was a vector and now it is a factor. Again, I wonder if old scripts need to be changed?