[Bioc-devel] [BioC] lookUp for other terminologies

Marc Carlson mcarlson at fhcrc.org
Mon Mar 7 19:56:31 CET 2011


Hi Matthew,

I think that you could probably do something like what you are
describing by following the instructions in the Vignette titled
"Hypergeometric tests for less common model organisms" which can be seen
in the GOstats package: 
http://www.bioconductor.org/help/bioc-views/release/bioc/html/GOstats.html. 
It was developed for a similarly generic circumstance as this.  An
extension of the methods developed for this case would be one way to
tackle this.

In terms of DO to gene mappings.  Where did you get them?  I have not
found these from any of the expected places.  If we can just get the
appropriate mappings from a reliable source then it might make sense to
include them in our packages.  However, if this is not practical, then
an approach like what was defined for unsupported organisms is probably
a good idea.


  Marc





On 03/07/2011 10:10 AM, Matthew Pocock wrote:
> On 7 March 2011 16:57, Martin Morgan <mtmorgan at fhcrc.org> wrote:
>
>   
>> I wonder if you could be a bit more specific about the tasks you're
>> trying to accomplish?
>>     
>
> There are lots of packages in bioconductor that provide methods that run
> with GO - GOstats, globaltest, GlobalAncova, ... and all of these that I've
> looked at so far essentially only need an annotation Bimap, some way to get
> terms, and something with the CHILDREN/OFFSPRING/... methods. I'd like to be
> able to run them for DO, but I'm sure there are other interesting
> terminologies out there that other people may want to use.
>
> I've been given an entrez<->do mapping file - I can't attest to its
> provenance or accuracy. From this, I've got functions that produce
> ${annotation}DO bimaps if there is an ${annotation}ENTREZID bimap present.
> As for the algorithms themselves, so far I've been copying out bits of these
> packages and hacking them about to use DO rather than GO* hierarchies.
> However, it's error-prone and not pretty, and I'd prefer to be able to do
> this with less hackery - or at least, by relying upon bioconductor-blessed
> and encapsulated hackery.
>
> Matthew
>
>



More information about the Bioc-devel mailing list