From GO Public
Inferences based on MF-BP Inter-Ontology Links
Since MFs are part_of some BP, the ontology editors have been adding some links between F and P aspects to connect the two aspects and give some context to the function. A simple example is: kinase activity part_of phosphorylation. Part_of links have been put in place in the ontology file (gene_ontology_write.obo file) between the MF and BP terms.
You can find some information in the wiki-
Once these links are created in the ontology file, Chris's script will take all the existing GAFs files and use these links to fill gaps by creating BP annotations if they are missing.
About the Inferrencing script
- if a geneproduct is annotated to protein kinase activity with evidenceCode1/paper1, but is missing a phosphorylation BP annotation with the same evidence code/reference, then the script will infer a phosphorylation BP annotation using the same evidence code and PMID as the MF annotation, and with 'GOC' in the assigned_by column.
- the inferencing script considers the term-evcode pair (and not reference) as an unique annotation, i.e., if a gene product annotated to Kinase activity with IDA has an annotation to phosphorylation but with IMP, it will create another annotation with the same evidence code as the MF annotation.
- if the gene product has an annotation to a child term of the BP term that is linked to the MF term, but with the same evidence code as the MF term, then the inference is not created
- if the gene product has an annotation to a child term of the BP term that is linked to the MF term, but with different ev.code, then an annotation to that BP term is inferred.
These inferences are available on the GO CVS web- http://cvsweb.geneontology.org/cgi-bin/cvsweb.cgi/go/scratch/gaf-inference/