Skip to main content
v2026.11,986 entries · CC-BY 4.0

GO Terms: Gene Ontology Aspects, IDs and Evidence Codes

What GO terms are: the three Gene Ontology aspects, how GO IDs work, evidence codes, annotations, enrichment backgrounds and how to cite GO.

Written and maintained by CASRAI Editorial Board

Published

On this page

GO terms are the standardized labels of the Gene Ontology (GO), each identified by a unique seven-digit number prefixed GO: (for example GO:0005739, mitochondrion). They describe what gene products do, in three separate aspects: molecular function, biological process and cellular component. Researchers attach GO terms to genes and proteins as annotations, then use them to summarize experiments, most often through GO enrichment analysis.

What are GO terms?

The Gene Ontology Consortium maintains GO as a structured vocabulary. A GO term is one node in that vocabulary: it has an identifier, a name and a definition, and it is linked to other terms by defined relationships. Because the identifier is stable and the vocabulary is shared, a result from one organism, laboratory or database can be compared with another without relying on free-text descriptions.

A GO term is not a gene and not a measurement. It is a category that a gene product can be annotated to. The link between a gene product and a term is a separate object, an annotation, which carries its own evidence and reference (covered below). Keeping these two ideas apart removes much of the confusion around reading GO results.

How to read a GO ID

According to the GO documentation, every term has a unique seven-digit identifier prefixed with GO:, such as GO:0005739 for mitochondrion. The digits carry no biological meaning you can decode: the number is simply a stable key. Quote the full identifier together with the term name in methods and supplementary tables, because the name alone can be ambiguous and the number alone is unreadable to a human reviewer.

The three GO aspects: molecular function, biological process and cellular component

GO is organised into three aspects, sometimes called namespaces or ontologies. The GO documentation defines them as follows.

Aspect Abbreviation What a term describes Annotation relation (GO annotation documentation)
Molecular function MF Molecular-level activities performed by gene products. These are activities, not the entities that perform them. enables
Biological process BP Larger biological programs accomplished by multiple molecular activities acting in a coordinated way. involved in
Cellular component CC The cellular location where a molecular function takes place. located in; part of

Molecular function (MF)

An MF term names an activity of a gene product, at the level of a single molecule or complex. The key point from the GO documentation is that MF terms describe the activity itself, not the protein or complex that carries it out. A gene product that has the activity is annotated with the relation “enables”.

Biological process (BP)

A BP term describes a larger program that results from several molecular activities working together. A gene product that contributes to such a program is annotated with “involved in”. BP is also the default aspect in the GO enrichment tool, so many published GO enrichment tables are, unless stated otherwise, tables of BP terms.

Cellular component (CC)

A CC term gives the location where a molecular function takes place. Gene products are annotated with “located in” or, for membership of a complex or structure, “part of”.

How GO terms relate to each other

Terms form a graph rather than a flat list. The basic link between terms is “is a”: a more specific term is a kind of a more general one. The documentation also describes the relations “part of” and “occurs in”, which can connect terms across aspects. This matters in practice because an annotation to a specific term implies the more general terms above it, which is why a gene annotated to a narrow process can still appear in a result for a broader one.

Some terms carry taxon constraints, meaning annotations to them are restricted to particular groups of organisms. If a term seems implausible for your species, the constraint is a first place to look.

GO annotations: connecting a gene product to a term

The GO annotation documentation describes an annotation as a statement made of four parts: a gene product, a GO term connected by a relation from the Relations Ontology (RO), a reference (a PMID, a DOI or a GO_REF identifier), and an evidence code. An annotation is called automated when it has not been reviewed by a biocurator.

Two further features are worth knowing. First, annotations can use the NOT modifier, which records that a gene product was experimentally shown not to have the stated relation to a term; the negation propagates to child terms. Second, annotations can be refined with annotation extensions, which also use RO relations. Annotation files are distributed in GAF 2.2 (the recommended format) and in the paired GPAD 2.0 and GPI 2.0 formats.

GO evidence codes

The evidence code on an annotation tells you how the link between gene product and term was established. This is the part of a GO result most often skipped, yet it decides how far the annotation can be trusted for a given question. The GO evidence-code guide groups the codes into categories, summarized in the data table below. The codes most readers meet are these:

  • IDA (inferred from direct assay), IPI (physical interaction), IMP (mutant phenotype), IGI (genetic interaction) and IEP (expression pattern) are experimental codes, alongside the general EXP.
  • HTP, HDA, HMP, HGI and HEP are the high-throughput counterparts of the experimental codes.
  • IBA, IBD, IKR and IRD are phylogenetic codes, inferring function from evolutionary relatives.
  • ISS, ISO, ISA, ISM, IGC and RCA are computational codes (sequence or structural similarity, orthology, alignment, sequence model, genomic context, reviewed computational analysis).
  • TAS and NAS are author statements; IC and ND are curator statements.
  • IEA (inferred from electronic annotation) is the one category that is not manually reviewed and is usually not traceable to an experimental source.

A practical consequence follows from the IEA definition. A very large gene list annotated mainly through electronic annotation will produce enrichment results that rest largely on unreviewed inference. Most enrichment tools let you filter by evidence code; whether to exclude IEA depends on your question, and either choice should be stated in the methods.

GO enrichment analysis: what the statistics ask

The most common use of GO terms is enrichment analysis. The GO enrichment page describes it as a way to find terms that are over- or under-represented in a gene set. For each term, the reported p-value is the probability of seeing at least x genes annotated to that term in a list of n genes, given the proportion of genes in the genome annotated to the term.

Choose the background list deliberately

The default background is all protein-coding genes of the genome, but the GO documentation strongly recommends a custom reference list: the genes that could actually have appeared in your result, such as all genes detected in your assay. This choice changes the p-values, because “over-represented” is always relative to the background. An experiment that measures only a subset of the genome (a targeted panel, or tissue-expressed genes only) will show spurious enrichment against a whole-genome background. State the background in your methods.

The enrichment tool defaults to the biological process aspect and the species Homo sapiens, and runs through PANTHER. The documentation also lists other tools, among them BiNGO, GeneWeaver, g:Profiler, GOrilla and Ontologizer. On this site, see the guides to g:Profiler, clusterProfiler and Enrichr for tool-specific workflows, and KEGG pathways for a complementary pathway resource. To look up the gene identifiers that feed these tools, see the NCBI Gene guide.

Common mistakes when using GO terms

  • Treating a term as a gene property. A term is a category; the annotation, with its evidence, is the claim.
  • Ignoring the default background. The default is the whole genome of protein-coding genes; the documentation recommends a custom list.
  • Mixing aspects. BP, MF and CC enrichment answer different questions. Report which aspect each table covers.
  • Not recording the version. GO changes between releases; results can differ for the same gene list.
  • Reading an annotation without its evidence code. IEA and experimental codes are not equivalent.

How to cite GO terms

The GO citation policy asks users to cite the original GO paper, Ashburner et al. (2000), Nature Genetics, DOI 10.1038/75556, and the Gene Ontology Consortium’s “The Gene Ontology knowledgebase in 2026” in Nucleic Acids Research, DOI 10.1093/nar/gkaf1292. Cite the tool you used as well: for example, PANTHER (Thomas et al. 2022, Protein Science, DOI 10.1002/pro.4218) or AmiGO (Carbon et al. 2009, Bioinformatics, DOI 10.1093/bioinformatics/btn615). The policy also asks you to state the GO release date or version used (its example is a release dated 2024-01-17, DOI 10.5281/zenodo.10536401) and to use PURLs for term links. GO content is released under CC BY 4.0.

Frequently asked questions

What does GO stand for in GO terms?

GO stands for Gene Ontology, the controlled vocabulary maintained by the Gene Ontology Consortium. Its terms describe gene product functions, processes and locations.

What are the three aspects of GO?

Molecular function (activities of gene products), biological process (larger programs carried out by coordinated activities) and cellular component (where a molecular function takes place).

How is a GO ID formatted?

A unique seven-digit number prefixed GO:, for example GO:0005739 for mitochondrion.

What is the difference between a GO term and a GO annotation?

A term is a vocabulary entry. An annotation links a gene product to a term through a relation, a reference and an evidence code.

Does IEA mean an annotation is wrong?

No. IEA marks annotations made electronically without manual review, usually not traceable to an experimental source. It is a statement about how the annotation was made, not a verdict on its correctness.

Which background should I use for GO enrichment?

The default is all protein-coding genes of the genome, but the GO documentation strongly recommends a custom reference list of genes that could have appeared in your result, such as all genes detected.

Sources

Data: GO evidence-code categories and the evidence codes in each

Compiled from the Gene Ontology Consortium evidence-code guide: each evidence category, the codes listed in it, the number of codes, and whether the guide describes the category as manually reviewed. Counts are the number of codes listed per category.

GO evidence-code categories and the evidence codes in each
Evidence categoryCodesNumber of codes
ExperimentalEXP, IDA, IPI, IMP, IGI, IEP6
High-throughputHTP, HDA, HMP, HGI, HEP5
PhylogeneticIBA, IBD, IKR, IRD4
ComputationalISS, ISO, ISA, ISM, IGC, RCA6
Author statementTAS, NAS2
Curator statementIC, ND2
ElectronicIEA1

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about GO Terms: Gene Ontology Aspects, IDs and Evidence Codes

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →