mcarlson at fhcrc.org
Sat Sep 19 01:38:01 CEST 2009
If your objective is to make a custom annotation package using
AnnotationDbi, and a mapping that relates some probes to genes, then the
SQLForge vignette in the AnnotationDbi package should tell you how to
proceed. You can find some information on that here:
If that is not what you wanted, then please let me know.
Tim Triche wrote:
> Sort of related-ly,
> Is there a good HOWTO type of document for packaging up Illumina
> annotation (CSV files/whatever) using AnnotationDbi? I don't have any
> problem parsing the file formats, but the vignette provided seems a
> bit Affy-centric and hard for me to adapt. I'm using R-devel and
> bioconductor-2.5 for assorted reasons.
> The methylation chips in particular require a little bit of chicanery,
> but that's not the problem; I'm having some issues figuring out how
> and what to put into which function so that other Bioconductor
> packages will know about the symbol, accession, etc. of the site being
> interrogated. I wrote a bit of a hack so that 'beadarray' could
> identify control probes and M vs. U probes on Infinium, but it's ugly
> and it doesn't help after summarization into an ExpressionSetIllumina
> or NChannelSet. I have the annotation information that I want to feed
> AnnotationDbi (as well as various masks, etc.) but I'm not sure how to
> structure it.
> There are a lot of (biological or sequence-related) features that can
> be used for probe masking on these arrays, of course, and I have been
> using several of them in an ad-hoc fashion (columns in a data frame
> with IlluminaIDs as rownames, excluding rows from the exprs matrix).
> I suspect there's a better way to do this, but (for example) the
> 'methylumi' and 'lumi' packages don't help me in this endeavor. Is
> there a customary place to store and operate upon probe masks, and is
> it separate from the annotation.db packages? Is there a good role
> model in terms of inheriting from a mask-able eSet class?
> Any assistance would be appreciated -- I'm sure these problems have
> been addressed before, but I just couldn't figure out where. Much
> obliged for anyone's time who cares to reply.
> On Fri, Sep 18, 2009 at 12:11 PM, Marc Carlson <mcarlson at fhcrc.org
> <mailto:mcarlson at fhcrc.org>> wrote:
> Hi Denise,
> Unfortunately, I the answer to your question is no. The annotation
> packages have had to undergo some pretty major schema changes in devel
> which means that they will necessarily be incompatible with the older
> version of AnnotationDbi that you will find in release. So you will
> either have to use R-devel or wait a month for R-devel to become the
> release version if you want to use the newer packages. In general, I
> would recommend against mixing packages from different releases as
> are not built or tested to work properly in this context.
> Therefore, I
> would also recommend against trying to install the newer AnnotationDbi
> in the older version of R.
> For your specific case, I think you should wait a month to upgrade
> annotation packages if you can do so. All the annotation packages are
> being refreshed as we speak so that they can be current for the
> release of Bioconductor. So after the new release should be an ideal
> time to upgrade. If you can't wait that long, then you can use
> with biocLite() to grab the latest packages.
> Denise Mauldin wrote:
> > Hey all,
> > >From other posts on the list it appears that AnnotationDBI has
> been updated
> > for the most recent R-devel. However, I'm using the 2.5
> metadata packages
> > to get the best updated information for my databases and the
> hgu133a.db (and
> > other affy) annotation packages require an AnnotationDBI that is
> higher than
> > 1.7.8. Is it possible to use these packages without upgrading
> to R-devel?
> > I currently have R 2.9.1 installed and I don't really want to go
> to a devel
> > version. Will AnnotationDBI 1.7.13 work with R 2.9.2 or is
> there another
> > way around this?
> > Thanks,
> > Denise
> > [[alternative HTML version deleted]]
> > _______________________________________________
> > Bioconductor mailing list
> > Bioconductor at stat.math.ethz.ch
> <mailto:Bioconductor at stat.math.ethz.ch>
> > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > Search the archives:
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch <mailto:Bioconductor at stat.math.ethz.ch>
> Search the archives:
> Statisticians, like artists, have a bad habit of falling in love with
> their models.
> --George Box
More information about the Bioconductor