Guidelines for electronic annotation methods
Electronic GO annotation
Electronic annotation is the process of assigning GO terms to gene products using automated methods. This process can rapidly produce millions of annotations in a very short period of time. The GO terms used for electronic annotation tend to be higher-level and, as such, less detailed than the terms used for manual annotation. It is possible to use certain electronic annotation methods if you have a new genome sequence to annotate, or a microarray with many thousands of sequences. See Using electronic annotation methods to map GO annotation to datasets
Annotations made using electronic annotation methods have the Inferred from Electronic Annotation IEA evidence code.
Types of electronic annotation used by the GO Consortium
Mappings to controlled vocabularies
One of the primary methods of generating electronic annotations is to manually map GO terms to corresponding concepts in the controlled vocabularies used by the UniProt Knowledgebase or other external resources. The mappings that are currently used by the GO Consortium to produce annotations are:
- Swiss-Prot keywords (SPKW2GO)
- Subcellular locations (SPSL2GO)
- Enzyme Commission numbers (EC2GO)
- InterPro domains (InterPro2GO)
- HAMAP (HAMAP2GO)
Projection of annotations between orthologous gene products
This method uses orthology data from Ensembl Compara to project GO annotations from a source species onto one or more target species.
Using electronic annotation methods to map GO annotation to datasets
Mapping files consist of entities from external databases indexed to identical, similar or related GO terms. Various mapping files have been created which can be used to provide novel sequences, ESTs, genomes sequences, microarray datasets or peptide sequences with GO annotation.
One of the most common mappings used to apply electronic GO annotation to user sequences is InterPro2GO.
It is possible to assign electronic GO annotation to your sequences based on InterPro domains. For example if your sequence has a DNA binding domain then it would be appropriate to electronically annotate it to the DNA binding function term. For more information on InterPro mapping please see the InterProScan information
Using sequence similarity to provide GO annotation to datasets
You can also provide your sequences with annotations by BLASTing them against manually annotated sequences and transferring the GO annotations across to your sequence. Since the originating GO annotation is experimentally derived, the annotation transferred to your sequences could have the Inferred from Sequence or Structural Similarity ISS evidence code.
Two tools which can assist you with this are;
Both of these tools allow you to choose the BLAST parameters, the values of which will depend on your individual requirements.
Perhaps the name for this page should change from electronic -> automatic annotation methods? As then you could include the f2p inferences - these won't have the IEA evidence tag (which seems to be implied by the 'electronic' wording of the page title), but are applied automatically from the f2p links in the GO file and it would be nice to have an explanation of the process here, especially as no GO_REF will be associated with the annotation set. I've drafted an explanation of this annotation method:
GOC Inferred Annotations
Annotations automatically generated using the Molecular Function->Biological Process inter-ontology relationships present in the GO OBO v1.2 format. As many GO users do not currently reason over these relationships, a set of inferred annotations are being generated. Such GO annotations are produced when an annotation has been made (either manually or electronically) to a Molecular Function term that, either directly or via one of its parent terms, has an relationship to a Biological Process term, and where this Process term (or one of its children) has not already been used in the annotation set for the same gene product identifier. This inferred annotation set applies the same gene product identifier, reference and evidence code as the asserted function annotation and are generated from all sources of GO annotations, with only 'NOT'-qualified annotations being excluded.
As an example, you can see an inter-ontology relationship exists to indicate 'GTPase activity' is a part_of the 'GTP catabolic process', as displayed here: http://www.ebi.ac.uk/QuickGO/GTerm?id=GO:0003924#term=ancchart . Edimmer