Annotation Extension: Difference between revisions

From GO Wiki
Jump to navigation Jump to search
Line 100: Line 100:


For some relationships, top-level domain restrictions apply, and in these cases then top-level terms from the Basic Formal Ontology (BFO) are used:
For some relationships, top-level domain restrictions apply, and in these cases then top-level terms from the Basic Formal Ontology (BFO) are used:


'''domain: BFO:0000007 ! process'''  
'''domain: BFO:0000007 ! process'''  

Revision as of 06:34, 1 August 2012

Introduction

Each GO annotation pairs a single gene product identifier to a single term from the ontology. This format is very powerful however it can also restrict the descriptiveness of a specific instance of a functioning or subcellular location; there must be a pre-existing term in the ontology that provides full details of the specific aspects of the function.

It is not always possible to create individual terms that precisely describe the context of each activity (e.g. the cellular or anatomical location, the dependency on other processes, or particular, specific protein targets).

It is less restrictive if the annotator is able to combine additional terms in a single annotation to provide a more detailed functional description for an individual gene product.

This page describes the Annotation Extension field (column 16) in the Gene Association File GAF2.0 file format, which allows GO terms to be further specified, using gene product or chemical identifiers or terms from GO or external OBO ontologies.

When an annotator chooses to do this, they are effectively creating "on-the-fly" cross-product term. We say "on-the-fly" because the combinatorial term is not added to the ontology (although it could be at a later stage, if the ontology editors choose to create the appropriate GO term).

When should a curator use the Annotation Extension field instead of requesting a new GO term?

The primary way to provide more fine-grained annotations is by requesting more specific terms. For example, it is reasonable to create sub-types of the general term “apoptotic process”, for example “anoikis” (definition: 'Apoptosis triggered by inadequate or inappropriate adherence to substrate e.g. after disruption of the interactions between normal epithelial cells and the extracellular matrix').

Terms should be requested via the ontology SourceForge tracker or via TermGenie.

However, highly detailed terms can inflate the ontology and lead to a GO term creation bottleneck. In addition it can be more efficient for an annotator to create the entire annotation statement in one step during the curation of the paper rather than in two steps that require making a term request and then going back later to include the new GO term in the annotation of the paper.

GO editors will regularly review the contents of the annotation extension field in submitted annotation files and create new, more specific terms if they feel enough annotations exist to warrant a pre-composed term. This effort will be assisted in future by automated methods to reason over annotations enhanced with filled annotation extension fields, to ensure the annotations are consistently grouped by an appropriate common GO term class.


1. GO term requests should not be made when the curator would like to describe activities or locations that are not evidently mechanistically or compositionally distinct from an existing GO term

Example

Rejected GO term: negative regulation of tRNA transcription from RNA polymerase III promoter

There is no specific negative regulator for tRNA transcription which is separable from general negative regulation of transcription from RNA polymerase III (where there is one highly conserved general negative regulator that shuts down all RNA polymerase III transcription).

Curators should therefore be advised to use the existing term "negative regulation of transcription from RNA polymerase III promoter" (GO:0016480) and capture any further specifics (such as the SO identifier for tRNA) in the annotation_extension field.

2. Term requests should not be made for specific extensions which are outside the scope of GO.

  • Specific protein substrates or products of an enzyme

Example:

Gene Name (col 2) GO ID (col 5) Reference (col 6) Annotation Extension (col 16)
NEK GO:0004672 protein kinase activity PMID:10880350 has_direct_input(UniProtKB:P36873 PPP1CC)
  • Specific chemical substrates for a catalytic or transporter activity are often considered trivial, therefore curators are recommended to discuss the appropriateness of a new term with the ontology editors.
  • N.B. The text colored green in the annotation examples on this page is not present in the annotation file, but is used here to improve the readers' understanding of annotation examples.

The basic format

An annotated GO term can be enhanced in the annotation format by one or more 'relationship(identifier)' pairs added into the annotation extension field (column 16).

The aim of the information added into the annotation extension field is to refine the GO term identifier entered into the GO_ID field (Column 5) of the annotation file.

For example, if a gene product Slp1 is localized to the plasma membrane of T-cells, the Gene Association File (GAF) would look like this (most columns omitted for brevity):

Gene Name (col 2) GO ID (col 5) Reference (col 6) Annotation Extension (col 16)
Slp1 GO:0005886 plasma membrane PMID:1234567 part_of(CL:0000084 T cell)

Here CL:0000084 is the identifier for T-cell in the OBO Cell Type (CL) Ontology.

All relationships used in the annotation extension field should be valid relationships (i.e. not obsolete) described in the go_annotation_extension_relations.obo file.

Identifiers can originate from GO or another ontology or database. All identifiers must be prefixed by the appropriate database namespace, e.g. UniProtKB:, CHEBI:, CL:, UBERON: and all database namespaces must be located in the GO Database Abbreviations file.

Annotation Extension Format:

relationship_type(database namespace:identifier)

Example:

requires_direct_regulator(UniProtKB:O43236)

Only one identifier can be referenced by one relationship.

Defining the appropriate relation(identifier) pair for an annotation using Domain and Range parameters

Relationship Domains

The Domain tag in a relationship's stanza refers to the type of GO identifier that has been present in GO_ID field, column 5 of GAF 2.0 of an annotation. Some relationships should not be used in annotations that use a GO identifier from a particular aspect or branch of the GO.

For some relationships, top-level domain restrictions apply, and in these cases then top-level terms from the Basic Formal Ontology (BFO) are used:

domain: BFO:0000007 ! process

Should be interpreted as a union of the Biological Process and Molecular Function ontologies

domain: BFO:0000001 ! entity

Should be interpreted as a union of Biological Process, Molecular Function and Cellular Component ontologies (i.e. all GO identifiers can be used)

domain: BFO:0000004 ! independent continuant

Should be interpreted as the Cellular Component ontology.

Relationship Ranges

The Range tag in a relationship's stanza refers to the types of database namespaces that can be appropriately used.

These Range values can be used as an additional check on the correctness of the relationship(identifier) pair.

In the Relations ontology, the range information is provided as named classes, identified by a ENTITY_UNION identifier.

Example:

range: ENTITY_UNION:0000003 ! gene or gene product

ENTITY_UNION identifiers are defined in separate stanzas in go_extension_rels.obo file:

id: ENTITY_UNION:0000003
name: gene or gene product
def: "The union of gene, RNA and protein entities." [GOC:ecd]
union_of SO:0000704 ! gene
union_of: SO:0000673 ! transcript
union_of: PR:000000001 ! protein 

Specific appropriate database identifiers can be identified using the 'entity_type' tags located in the GO References file:

abbreviation: UniProtKB
database: Universal Protein Knowledgebase
description: A central repository of protein sequence and function created by joining the information 
contained in Swiss-Prot, TrEMBL, and PIR databases
entity_type: PR:000000001 ! protein 

Examples of appropriate and inappropriate usage of relation(identifier) pairs

Many relationships have been restricted according to the type of term they can appropriately describe, and also the types of values that can be assigned.

Example from the go_annotation_extension_relations.obo file:

[Typedef]

id: has_regulation_target

name: has_regulation_target

def: "Identifies a gene or gene product affected by a regulation BP." [GOC:mah]

xref: GOREL:0000015

domain: GO:0065007 ! biological regulation

range: ENTITY_UNION:0000003 ! gene or gene product 

The above extract of the go_annotation_extension_relations.obo file indicates that the has_regulation_target relationship can only correctly be applied in annotations which have used a GO term in column 5 that is a descendant of 'GO:0065007 biological regulation'.

In addition, only gene or gene product identifiers should be the types of values associated with this relationship, as indicated by the 'range' tag.

Therefore, while this annotation is correct:


Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
DAOA GO:1900758 negative regulation of D-amino-acid oxidase activity PMID:21679769 IDA has_regulation_target(UniProtKB:P14920 D-amino-acid oxidase)

The following annotations, although containing appropriate information for the annotation extension field, inappropriately uses the has_regulation_target relationship as their use conflicts with the specified defined Range and Domain values defined in the description of the relation above.


Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
PEX19 GO:0072662 protein localization to peroxisome PMID:18782765 IMP has_regulation_target(UniProtKB:Q9Y3D6 Mitochondrial fission 1 protein)


The example above is not acceptable because the GO term is not a descendant of 'biological regulation'.

Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
C-C motif chemokine 24 GO:0008360 regulation of cell shape PMID:10072545 IDA has_regulation_target(CL:0000771 eosinophil)


The example above is not acceptable because cell type ontology (CL) identifiers are not included within the Range scope of the has_regulation_target relation.

  • N.B. The text colored green in the annotation examples on this page is not present in the annotation file, but is used here to improve the readers' understanding of annotation examples.


Format for referencing multiple relationship identifier pairs

Only one identifier can be referenced by one relationship, therefore in order to make multiple 'relationship(identifier)' pairs in an annotation_extension field they must be separated with commas ',' or pipes '|'.

Very simply, the pipe can be interpreted as meaning "or" and comma meaning "and".

Use of the Pipe to separate Annotation Extension values

The current annotation format guidelines states that as inclusion of data in the annotation extension field is entirely optional for the correct interpretation of an annotation, two annotations should not exist that only differ by the contents of their annotation extension field.

Therefore where a gene product carries out its activity in different places or under different circumstances, then multiple annotation extension 'relationship(identifier)' pairs should be added into the annotation extension field of the same annotation and be separated from each other with a pipe. This format indicates to the user that the different relationship(identifier) pairs are making completely independent statements, and it would be equally correct to represent the annotation extension data in separate annotation lines.

Example

Where a gene product can act catalytically on any of a number of different substrates, but not all at the same instance, then the different 'relationship(identifier)' pairs should be separated using pipes:

Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
Nos2 GO:0005777 peroxisome PMID:12085352 IDA

part_of(CL:0000182 hepatocyte)|part_of(CL:0000091 Kupffer cell)


Interpretation:

Nos2 has been obsvered by direct assay to be located in the peroxisome of hepatocyte cells, and also located in the peroxisome of Kupffer cells.

The annotation format using multiple, pipe-separated 'relationship(identifier)' pairs supplies the equivalent information as supplying two separate annotation statement lines, each with one of the different relationship-value pairs:


Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
Nos2 GO:0005777 peroxisome PMID:15127951 IDA part_of(CL:0000182 hepatocyte)
Nos2 GO:0005777 peroxisome PMID:15127951 IDA part_of(CL:0000091 Kupffer cell)

Use of the Comma to separate multiple Annotation Extension values

Commas enable curators to create "compound" annotation extensions. This format is applied where a combination of 'relationship(identifier)' pairs supplies a complex, detailed description of the context or specific nature of an instance of a function/location.

Example:

Gene Name (col 2) GO ID (col 5) Reference (col 6) Evidence (col 7) Annotation Extension (col 16)
TMEM11 GO:0005634 nucleus PMID: 17973242 IDA part_of(CL:0000066 epithelial cell),part_of(UBERON:0004801 cervix epithelium)

Interpretation:

TMEM115 is located in the nucleus that are part of epithelial cells (CL:0000066) that are part of the cervix epithelium (UBERON:0004801).

  • N.B. The text colored green in the annotation examples on this page is not present in the annotation file, but is used here to improve the readers' understanding of annotation examples.

Annotation Examples