Skip to content

[Bioc-devel] Bioc-devel Digest, Vol 145, Issue 60

5 messages · Aedin Culhane, Vincent Carey, Martin Morgan

Original
Aedin Culhane
# 
Hi
Using select/mapIDs to annotate probe IDs is an additional step that 
confuses many.

May I suggest we automatically populate fData with minimal annotation 
(ProbeID, entrez ID, symbol) if a known platform is detected.   We 
record the version and parameters (eg mapIDs multi=first) used to create 
fData.  But for beginners I think it would be a helpful start.


What do you think?
Aedin
Vincent Carey
# 
I am in favor of simplifying the binding of useful metadata to our
genome-wide objects.  Before we automate this I think we
should define a widely applicable procedure for this task ... and see how
it works in examples from the ExperimentData library and
ExperimentHub.  Using fData for ExpressionSet and rowData for
SummarizedExperiment and rowRanges for RangedSummarizedExperiment
might also be susceptible of simplification.  fAnno?

On Wed, Apr 20, 2016 at 4:10 PM, Aedin Culhane <aedin at jimmy.harvard.edu>
wrote:

  
  


  


      

    

    

      
      

        


  

        

      Aedin Culhane
    

    

      
      #
    

  


  

      
Hi Vince
Agreed, I agree fData is over-simplification. But if data have an 
associated "annotation", the feature annotation associated with it 
should be available.

I was recently trying to map between one of the hugene st1.0 and 
primeview arrays.  The first has multiple .db packages  (including 
hugene10stprobeset.db,
hugene10sttranscriptcluster.db) and its not very clear which is the 
correct one to use.   There is no .db package for primeview, so I had to 
download the .csv file from the Affy website and build the package.

The probe genome co-ordinates would allow better merging of platforms 
(as opposed to mapping identifiers to a common entrez gene id/transcript 
id). Moreover, with GRanges, we could use mapToTranscripts, 
findOverlaps, countOverlaps to map between platforms.

A.

        
On 4/20/16 22:20, Vincent Carey wrote:

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Martin Morgan
    

    

      
      #
    

  


  

      
On 04/20/2016 11:15 PM, Aedin Culhane wrote:

            

      
      
      
    

        
so auto-filling annotations wouldn't have helped here, because there is 
not an automatic choice between alternate packages and because the 
primeview array doesn't have an annotation package?

            

      
      
      
    

        
the brain array project makes it clear that there's more than one way to 
map a probe to a gene. All we do is report what the manufacturer says. I 
think we don't have the resources (technical expertise, in addition to 
labor) to re-inventing and maintain our own mappings.

Martin

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        
This email message may contain legally privileged and/or confidential information.  If you are not the intended recipient(s), or the employee or agent responsible for the delivery of this message to the intended recipient(s), you are hereby notified that any disclosure, copying, distribution, or use of this email message is prohibited.  If you have received this message in error, please notify the sender immediately by e-mail and delete this email message from your computer. Thank you.

  
  


  


      

    

    

      
      

        


  

        

      Vincent Carey
    

    

      
      #
    

  


  

      
On Thu, Apr 21, 2016 at 8:35 AM, Martin Morgan <

        
martin.morgan at roswellpark.org> wrote:

        

            

      
      
      
    

        
Yes.  I think there are two threads emerging here.

First, one focused on support for affy arrays.  We've lost momentum here
but may still be the system of first resort for people
who want to preprocess and analyze them.  More vignettes and benchmarking
for the newer releases like HTA and primeview
would probably be welcome to researchers who use them (120 GDS in GEO for
primeview, one of which cites affy/RMA;  82 GDS on HTA 2.0 gene version).
I think someone who is heavily invested in these platforms would have to
step up to fill gaps in our support.

Second, the concept of streamlining the attachment of annotation
information to array or sequence-based quantifications.
Here we need data from the community about current gaps and successes.  I
think it would be helpful (and feasible) to have
a generic that addresses this, but there is variation in the number of
resources to be consulted to annotate a given platform, and no natural
choice of resource for various types of feature.   So programming and
documentation at the user level seem inevitable for any given solution.
Centralized efforts may not pay off.