Hi all,

I've got a question concerning the GO.db package, in particular about the annotation of GO identifiers to their parents. I read the pdf that describes the functions of GOMFPARENTS, GOCCPARENTS and GOBPPARENTS and was wondering if there is a little mistake: At the GOBPPARENTS description it says "This data set describes associations between GO molecular function (BP) terms and [...]". Shouldn't it be "This data set describes associations between GO biological process (BP) terms and [...]". The same mistake(?) is at the description of GOCCPARENTS.

Then, there is another question:
I looked up the node "GO:0016301" at the gene ontology consortium website. There is a graph that shows, that this node has two parents: node number "GO:0016772" and number "GO:0016310". Now, if I do in R:
 

library(GO.db)
x<-as.list(GOMFPARENTS)
xx<-as.list(GOCCPARENTS)
xxx<-as.list(GOBPPARENTS)
x['GO:0016301']
#$`GO:0016301`
#       is_a
#"GO:0016772"
xx['GO:0016301']
#$<NA>
#NULL
xxx['GO:0016301']
#$<NA>
#NULL

I can't find the parent-node number "GO:0016310", which has -following AmiGO - a "part_of"- relation to the node "GO:0016301". And in the details of GO_BP/MF/CC_PARENTS it says, that "[t]he name associated with the, parent term will be either isa, hasa or partof". Hence, I thought the term "GO:0016310" should be in the data. I could imagine that it is a data base problem. Maybe GO.db isn't as up-to-date as AmiGO. Do you think this actually is the reason or is there another problem?

Thanks in advance,
1frcn

 version
#
#platform       x86_64-w64-mingw32         
#arch           x86_64                     
#os             mingw32                    
#system         x86_64, mingw32            
#status                                    
#major          3                          
#minor          1.1                        
#year           2014                       
#month          07                         
#day            10                         
#svn rev        66115                      
#language       R                          
#version.string R version 3.1.1 (2014-07-10)

I don't think that GO.db captures information across name spaces, such as the part_of that you describe or occurs_in from Annotation Extension for GO terms post. I think it would definitely be interesting to have such functionality in Bioc.

Hi,

So thank you for mentioning the typo in the documentation.  I will make an effort to tidy those up. 

It turns out that the missing term you mentioned ("GO:0016310") is not actually missing from the GO.db database.  You can test that idea out directly by using select like this:

library(GO.db)
select(GO.db, 'GO:0016310', columns = c('TERM','ONTOLOGY'), keytype='GOID')

So the reason it doesn't show up in the GOMFPARENTS mapping is because that mapping is an 'MF' mapping.  Whereas the "GO:0016310" term you were expecting to see as a parent is actually a 'BP' term. 

What is probably confusing you is the image from Amigo (drawn below).  And the reason it's confusing is that they are drawing all the different GO terms interlaced together as a single graph (when in fact GO is actually three separate ontologies).  If you look closely at the image you will notice that at the top you can see the roots of all three major ontologies represented...  In contrast, the MFPARENTS mapping that you were searching in can only get you parents from the 'MF' ontology.