Biology Atlas

Map / Concepts

concept · intermediate · partly written

Gene Ontology enrichment analysis, and how it fails

Enrichment analysis asks whether a gene list hits a GO term more often than chance, so its answer depends on the annotation table, the background and the release.

Also called: over-representation analysis, GO enrichment

How it works

Choose a gene list and a background (the genes that could have been in the list). Count how many list genes carry each term, test each term with a hypergeometric, chi-square or binomial test, and correct for the many terms tested. The statistics are simple; the answer depends on the annotation table behind the counts. 12

Established

Release instability more

Across 104 disease analyses, the median overlap (Jaccard index) of enriched terms with a 2015 reference rose from 0.27 in 2004 to 1 in 2015, and stayed between 0.038 and 0.1 until 2010 when the annotations were varied. It tracked the share of well-characterised genes (R = 0.984). The authors caution that 2015 is no gold standard and advise re-analysis; they do not show that old published results are wrong. 3

Reported, not yet replicated

Concentration of annotation more

Because 58% of annotations sat on 16% of human protein-coding genes in 2015, a term can come out enriched because the genes under it have been studied hard. Papers per gene are predicted by gene features with a Spearman correlation of 0.64, and 27% of genes never had a full publication. 34

Reported, not yet replicated

Electronic against experimental annotation more

In January 2015, 57% of human gene annotations were electronic (IEA), with no curator. The reliability of electronic annotation (the share later confirmed by experiment) was 0.52, against 0.33 for curated non-experimental annotation, so the electronic share is not simply a quality problem. The two figures use different denominators and years. 35

Reported, not yet replicated

The open-world problem more

Under 1% of experimental annotations in UniProt-GOA are negative, so a missing annotation means nobody has reviewed the gene, not that it has no function. This figure is the authors' statement and cites another paper. 6

Reported, not yet replicated

Gene length and other test assumptions more

For RNA-seq, longer genes are more likely to be called differentially expressed, which biases the count. The standard test also treats genes and terms as independent, discards fold change and relies on an arbitrary cutoff. 71

Reported, not yet replicated

What practitioners do about it more

Use a background that matches the experiment, treat or exclude electronic annotations, avoid protein-binding annotations in enrichment, correct for multiple testing, record the GO release date and repeat across several releases. These are recommendations by the Consortium and by the authors cited; this atlas found no study of how many published results change when a correct background is used. 86

Reported, not yet replicated

Plain answer

Enrichment analysis tests whether your genes are over-represented under a term. Where the test is trivial, the annotation table is not: it records where science has looked, so results move with the release and with which genes are well studied. 31

Reported, not yet replicated

Why it matters

This is the clearest measured case in this atlas where the data layer, not the algorithm, decides the answer. 3

Reported, not yet replicated

Real numbers

Median agreement of enriched terms, GO 2004 against 2015

Jaccard 0.27 (1 = identical). 3

Reported, not yet replicated

Share of human annotations that were electronic, January 2015

57%. 3

Reported, not yet replicated

Still open here

Not yet written: Equation.

Linked from

Sources

  1. Khatri P, Sirota M, Butte AJ. Ten years of pathway analysis: current approaches and outstanding challenges. PLoS computational biology (2012). paper; opened 2026-11-18 all records citing this
  2. GO enrichment analysis.. organisation-page, not peer reviewed; opened 2026-11-18 all records citing this
  3. Tomczak A, Mortensen JM, Winnenburg R, Liu C, et al.. Interpretation of biological experiments changes with evolution of the Gene Ontology and its annotations. Scientific reports (2018). paper; opened 2026-11-18 all records citing this
  4. Stoeger T, Gerlach M, Morimoto RI, Nunes Amaral LA. Large-scale investigation of the reasons why potentially important genes are ignored. PLoS biology (2018). paper; opened 2026-11-18 all records citing this
  5. Skunca N, Altenhoff A, Dessimoz C. Quality of computationally inferred gene ontology annotations. PLoS computational biology (2012). paper; opened 2026-11-18 all records citing this
  6. Pascale Gaudet, Christophe Dessimoz. Gene Ontology: Pitfalls, Biases, Remedies. arXiv (2016). preprint, not peer reviewed; opened 2026-11-18 all records citing this
  7. Young MD, Wakefield MJ, Smyth GK, Oshlack A. Gene ontology analysis for RNA-seq: accounting for selection bias. Genome biology (2010). paper; opened 2026-11-18 all records citing this
  8. Gene Ontology Consortium, Aleksander SA, Balhoff J, Carbon S, et al.. The Gene Ontology knowledgebase in 2023. Genetics (2023). paper; opened 2026-11-18 all records citing this
Where it sits on the map
History of this record (6)

This record has been changed 6 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.

Status
partly written
Last updated
2026-11-19
Record
content/concepts/go-enrichment-analysis.json