[Bioc-devel] Re: [BioC] can't instal MetaData package "mouse4302"

I've moved this to bioc-devel.

At 12:33 PM 13/03/2005, Robert Gentleman wrote:

Hi Gordon,
  Thanks for the careful report - we will be looking in to this and  
see
what can be resolved. Some issues are due to the move to Seattle
(hopefully that is mostly resolved now), and some due to the change
from LocusLink to EntrezGene - and that is more problematic as our
basic data sources have changed completely.

Thanks, that's great.

I'd really like to have a better understanding of how the annotation  
packages are put together, the critical thing being the mapping of  
Affy probe set IDs to Entrez Gene (Locus Link) IDs. There's some  
information about this at

  http://www.bioconductor.org/data/annotation.html

but it doesn't fully describe the process. I notice that you're not  
just accepting the Affymetrix supplied Locus Link IDs. In your paper  
with John Zhang,  
http://bioinformatics.oupjournals.org/cgi/reprint/19/1/155, you  
describe a process which accepts the Affymetrix supplied GenBank IDs,  
and then uses a voting of other data sources to map GenBank to Locus  
Link. Is that still the strategy? Is that better than just accepting  
the Affymetrix supplied LL IDs?

As far as I know - and things are changing rapidly - Affymetrix  
version of things is much better now than it was when we started (at  
least for standard arrays, I am not sure about custom ones) and we are  
going to start to make more use of their data. When we first started  
the answer to your last
question was yes, we were doing much better.

  In any event, Affymetrix does not give mappings to everything we want,  
and so we will still have to do some manipulations and in some cases  
there does seem to be reasonable disagreement (and it is nice to know  
about such cases) - however, as I said, for standard arrays things are  
pretty stable and we are generally only conflicting on fairly obscure  
things.

Would it be possible to include with each annotation package the code  
used to build it, say in a directory called /build or whatever? The  
code might just be a few calls ot AnnBuilder functions, but it would  
nice to document it.

Well, that is the point of AnnBuilder itself and we do document the  
actual resources and version numbers that have been used, for example  
go:
   ?hgu95av2
and that is pretty explicit. We can look at adding more, but I doubt  
that we can easily give something that is going to be comprehensive.  
The data resources themselves can undergo substantial changes (which  
seems to be what is happening with the mouse data right now) and we are  
struggling to find enough resources to keep up with those changes. Our  
main priority is ensuring that what we produce is accurate - so we do  
appreciate your comments and observations and will try to get things  
running more smoothly.

   I hope that from now on we will be able to provide some basic  
summaries on what has changed when we produce new sets of annotation  
(but again, getting that to work in anything that looks like an  
automatic fashion is non-trivial).

   Robert

Cheers
Gordon

  Best wishes,
    Robert

On Mar 11, 2005, at 10:40 PM, Gordon Smyth wrote:

At 01:11 PM 12/03/2005, Seth Falcon wrote:

Hi Gordon,

The mouse4302 annotation package is again available.  The version is
1.6.9.  The data is the same as version 1.6.8, the packaging was
modified so that the package is usable on Windows.

Sorry for the delay.
Best,
+ seth

Hi Seth,

Many thanks for this.  I'll update my recent Mouse 430 2.0 analyses
with this package.

I've compared versions 1.6.5 and 1.6.9 of the mouse4302 annotation
package and found somewhat more differences than I expected.

...

+----------------------------------------------------------------------- 
----------------+
| Robert Gentleman              phone: (206) 667-7700                    
          |
| Head, Program in Computational Biology   fax:  (206) 667-1319   |
| Division of Public Health Sciences       office: M2-B865               
       |
| Fred Hutchinson Cancer Research Center                                 
          |
| email: rgentlem@fhcrc.org                                              
                          |
+----------------------------------------------------------------------- 
----------------+

A couple of clarifications:

Some issues are due to the move to Seattle (hopefully that is
mostly resolved now), and some due to the change from LocusLink to
EntrezGene - and that is more problematic as our basic data sources
have changed completely.

The mouse4302_1.6.9 annotation package that is currently on
www.bioconductor.org was created, without any change to the data, from
version 1.6.8.  The 1.6.9 version resolves an issue with data file
packaging that prevented use of the package on Windows.

Version 1.6.8 was built in early January 2005 by Jianhua (according to
the DESCRIPTION file). My understanding is that this package would
have been built against LocusLink and not EntrezGene.

+ seth