Biology Atlas

Map / Concepts

concept · intermediate · partly written

How entity grounding works and fails in knowledge graphs

Grounding turns a string such as "ERK" into one identifier; most graphs publish no accuracy for this step, and the documented failures are shared aliases, wrong species and family versus member.

Short answer

Grounding maps a text string to one database identifier, at reading time and again when databases are merged. Measured numbers exist only for the text-reading tools. PrimeKG and RTX-KG2 report no precision or recall for their identifier mapping. 1234

Reported, not yet replicated

Lexicon lookup and ambiguity more

Gilda matches strings against over 1.8 million names from 9 resources. 3,662 synonyms are shared by at least two human genes and 569 by three or more; DAP4 is a synonym of both DLGAP4 and THAP12. On BioCreative Bio-ID, overall F1 was 0.618 and human-gene F1 0.694 (0.762 if family matches count, since the benchmark labels VEGF as VEGFA while Gilda returns the family). Context models moved overall F1 only from 0.614 to 0.618. These are the authors' own runs. 1

Reported, not yet replicated

Families and complexes more

FamPlex, a curated table of 441 families and complexes covering 2,040 genes, raised correct grounding of families and complexes from 15.1% (8 of 53 sampled) to 71.2% (37 of 52). Without it "ERK" went to EPHB2 and "AKT" to a slime-mould gene. Samples are small. 2

Reported, not yet replicated

Mapping layers more

Biomappings holds 9,274 human-confirmed true mappings and 1,215 confirmed false, with measured prediction precision of 97 to 98% for chemicals (a sample of 100) and 63 to 68% for Cellosaurus cell lines; it flags "Severe Dengue" (MeSH) and "Dengue Shock Syndrome" (UMLS) as wrongly equivalent. SemRep's error analysis puts 26.9% of its errors in entity recognition and normalisation, at strict precision 0.55 and recall 0.34. 56

Reported, not yet replicated

Damage upstream of any graph deep

One fifth of papers with Excel supplementary gene lists had gene names converted to dates or numbers (abstract level read). 7

Reported, not yet replicated

Still open here

Not yet written: Plain answer, Why it matters, Equation, Real numbers.

Linked from

Sources

  1. Gyori BM, Hoyt CT, Steppi A. Gilda: biomedical entity text normalization with machine-learned disambiguation as a service. Bioinformatics Advances (2022). paper; opened 2026-11-18 all records citing this
  2. Bachman JA, Gyori BM, Sorger PK. FamPlex: a resource for entity recognition and relationship resolution of human protein families and complexes in biomedical text mining. BMC Bioinformatics (2018). paper; opened 2026-11-18 all records citing this
  3. Chandak P, Huang K, Zitnik M. Building a knowledge graph to enable precision medicine. Scientific Data 10:67 (2023). paper; opened 2026-11-18 all records citing this
  4. Wood EC, et al.. RTX-KG2: a system for building a semantically standardized knowledge graph for translational biomedicine. BMC Bioinformatics (2022). paper; opened 2026-11-18 all records citing this
  5. Hoyt CT, Hoyt AL, et al.. Prediction and curation of missing biomedical identifier mappings with Biomappings. Bioinformatics (Oxford, England) (2023). paper; opened 2026-11-18 all records citing this
  6. Kilicoglu H, et al.. Broad-coverage biomedical relation extraction with SemRep. BMC Bioinformatics (2020). paper; opened 2026-11-18 all records citing this
  7. Ziemann, M., Eren, Y., El-Osta, A.. Gene name errors are widespread in the scientific literature. Genome Biology 17 (2016). paper; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (5)

This record has been changed 5 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/entity-grounding-in-knowledge-graphs.json