Seth, Thanks for your offer of help on this. I am very much a novice compared to you, but I would be happy to help with coding to move things along at some point. This is not an urgent issue for me, as we are going to submit a grant in a week, so will just outline our methods for now. Mark Mark W. Kimpel MD (317) 490-5129 Work, & Mobile (317) 663-0513 Home (no voice mail please) 1-(317)-536-2730 FAX -----Original Message----- From: Seth Falcon [mailto:sfalcon@fhcrc.org] Sent: Monday, January 22, 2007 12:06 PM To: Kimpel, Mark William Cc: bioconductor at stat.math.ethz.ch; Seth Falcon Subject: Re: hyperGTest on proteomics data "Kimpel, Mark William" <mkimpel at="" iupui.edu=""> writes: > I would like to use the hyperGTest in package "GOstats" on some > proteomics data that I have. It is LCMS data and I have a list of all > Entrez Gene identifiers of all identified proteins (the universe) and of > the significant proteins. > I imagine this is because hyperGTest relies on the Bioc metadata > packages to map EntrezGene to GO categories. It seems to me that there > should be a work around to this problem using biomaRt or some direct > calls to the GO website. Your assessment is correct. The code in Category and GOstats relies upon environments from an annotation data package to retrieve the required mappings. It should be possible to obtain the mappings required for the test using just the GO package, but it will require some work... > Suggestions? I think the way to go is to add either a new parameter class and/or a new class for representing the annotation data (in devel there is already code along these lines to make use of SQLite-based annotation packages). Then one would need to define methods for universeBuilder and categoryToEntrezBuilder that extract the required info just using the GO package. At that point, things should plug into the rest of the code and Just Work (yeah, sure). It is something that is on my TODO list, but I don't have an ETA for when it will be done. + seth