Gene Ontologies annotation incomplete in Biomart?

[Joe Dougherty]

Hi,

I was querying biomart through R to try and pull down all ensemble gene IDs associated with a particular GO term. - in this case muscle development

e.g.:

ensembl = useMart("ensembl",dataset="hsapiens_gene_ensembl") 

gene.data <- getBM(attributes=c('hgnc_symbol', 'ensembl_gene_id', 'go_id'), filters = 'go_id', values ="GO:0060537", mart = ensembl)

And I noticed this would only return one gene for me.  However, looking at the term GO:0060537 at the Amigo portal returns a much larger number of homo sapiens genes.

I next checked at the biomart webportal, again just querying with GO:0060537, and again, it only returns one gene (Srpk3).  Does Biomart not contain the full GO annotation? Or is there some fluke just with this one GO term?

Joe

[Arek Kasprzyk]

Hi Joe,

Quick lookup on the Ensembl site with GO:0060537

http://www.ensembl.org/Homo_sapiens/Search/Results?q=GO%3A0060537;site=ensembl;facet_species=Human

returns a bunch of transcripts that belong to the SRPK3 gene. So it appears that BioMart Ensembl data is in sync with the Ensembl site.

If you need a clarification why Ensembl is not showing the same data as the Amigo portal, it is best if you contact Ensembl experts directly

at help...@ensembl.org

a

--
You received this message because you are subscribed to the Google Groups "biomart-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biomart-user...@googlegroups.com.
Visit this group at http://groups.google.com/group/biomart-users.
For more options, visit https://groups.google.com/groups/opt_out.

--

"In prosperity, our friends know us; in adversity, we know our friends"

― John Churton Collins 

[Thomas Maurel]

Dear Joe,

If you use the "go_id" filter, biomart will return all the genes linked to this GO term in ensembl. You can see on the amigo website that GO:0060537 is also only linked to the human gene SRPK3: http://amigo.geneontology.org/cgi-bin/amigo/term-assoc.cgi?term=GO:0060537&speciesdb=all&taxid=9606.

The other genes that you get from the amigo website are actually genes linked to children of the GO term GO:0060537. You can get these genes back in biomart by using the "go_parent_term" filter, this will return all the genes linked to GO:0060537 as well as the genes linked to the children terms.

gene.data <- getBM(attributes=c('hgnc_symbol', 'ensembl_gene_id', 'go_id'), filters = 'go_parent_term', values ="GO:0060537", mart = mart)

> gene.data[0:10,]
   hgnc_symbol ensembl_gene_id      go_id
1        FOXL2 ENSG00000183770 GO:0002074
2        PITX2 ENSG00000164093 GO:0002074
3      CACNA1S ENSG00000081248 GO:0002074
4         TBX3 ENSG00000135111 GO:0003167
5       NKX2-5 ENSG00000183072 GO:0003168
6         RYR2 ENSG00000198626 GO:0003220
7       NKX2-5 ENSG00000183072 GO:0003221
8        ZFPM2 ENSG00000169946 GO:0003221
9         DLL4 ENSG00000128917 GO:0003222

10       FOXH1 ENSG00000160973 GO:0003222

Hope this helps,

Thomas

--

Thomas Maurel
Bioinformatician - Ensembl Production Team
European Bioinformatics Institute (EMBL-EBI)
European Molecular Biology Laboratory
Wellcome Trust Genome Campus
Hinxton
Cambridge CB10 1SD
United Kingdom

[Bernard Thienpont]

Hi Thomas,

I am experiencing a similar issue, as I would like to annotate each gene also with the parents of GO terms associated with it (and not only to the direct GO term it is annotated with).

gene.data <- getBM(attributes=c('hgnc_symbol', 'ensembl_gene_id', 'go_parent_term'), filters = 'go_parent_term', values ="GO:0060537", mart = mart)
doesn't work because go_parent_term is not a list attribute. Would you happen to know a simple work-around?

thanks heaps in advance for your help - have been biting my teeth on this for a few hours now...

Bernard

[Thomas Maurel]

Dear Bernard,

I am afraid that this is not possible with the ensembl mart at the moment. The "go_parent_term" filter is actually using an other mart called "ontology_mart" in the background to get all the GO children terms. This is the reason why we don't have a "go_parent_term" attribute in the ensembl mart.

Regards,

Thomas

[Bernard Thienpont]

OK, thanks for your help.