Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

Quantifying Specificity of GO Terms

Paul Pavlidis paul at bioinformatics.ubc.ca
Fri Apr 20 10:52:50 PDT 2007


Just to throw our hat into the ring, we experimented with both the Resnik measure and a 
simple 'GO term overlap' measure and found that they are highly correlated. Measuring term 
overlap (the number of GO terms shared by two genes, including all parent terms) does not 
depend on knowing term-use frequencies and very fast to compute. The term overlap measure 
is used and briefly described here:

http://www.genome.org/cgi/content/full/14/6/1085

As for whether it is "more accurate", I don't know (I'm interested in hearing any 
opinions), but as I believe was pointed out, the depth of terms in the hierarchy is 
actually not a bad indication of their specificity.

Paul

Paul Pavlidis, PhD
Assistant Professor of Psychiatry
UBC Bioinformatics Centre (UBiC)
177 Michael Smith Laboratories
2185 East Mall
University of British Columbia
Vancouver BC V6T1Z4
voice: 604 827 4157
fax: 604 608 2964
paul at bioinformatics.ubc.ca
http://bioinformatics.ubc.ca/pavlidis/



Phillip Lord wrote:
 >
 > Probably because Lin and Jiang are both normalised measures, while
 > Resnik is not. My over-riding suspicion has been that this distinction
 > is more important than anything else.
 >
 >>>>>> "RA" == Rubio, Angel <arubio at ceit.es> writes:
 >
 >   RA> Some years ago, my group compared the correlation between gene
 >   RA> expression and different versions of semantic similarity. We
 >   RA> found it that the Resnik similarity measure (already used by
 >   RA> Dr. Lord comparing sequence and functions) outperformed other
 >   RA> measures also based on a corpus for the three categories (BP, MF
 >   RA> and CC).  Indeed, in our case these other measures (Lin and
 >   RA> Jiang) did not perform well at all.  Resnik similarity measure
 >   RA> is easy to evaluate:
 >
 >   RA> Resnik(GeneProduct1, GeneProduct2) = -log(ni/nt)
 >
 >   RA> Where ni: number of gene products in the corpus annotated for
 >   RA> the common ancestor of the annotations of a pair of gene
 >   RA> products (it seems a sort of tongue twister!).  nt: total number
 >   RA> of gene products.
 >
 >   RA> I expect that it helps.
 >
 > --
 > This message is from the GOFriends moderated mailing list.  A list of public
 > announcements and discussion of the Gene Ontology (GO) project.
 > Problems with the list?           E-mail: owner-gofriends at geneontology.org
 > Subscribing   send   "subscribe"   to   gofriends-request at geneontology.org
 > Unsubscribing send   "unsubscribe"  to  gofriends-request at geneontology.org
 > Web:          http://www.geneontology.org/


--
This message is from the GOFriends moderated mailing list.  A list of public
announcements and discussion of the Gene Ontology (GO) project.
Problems with the list?           E-mail: owner-gofriends at geneontology.org
Subscribing   send   "subscribe"   to   gofriends-request at geneontology.org
Unsubscribing send   "unsubscribe"  to  gofriends-request at geneontology.org
Web:          http://www.geneontology.org/



More information about the go-friends mailing list