Ontology · LiteFold
GO
A structured tabular Parquet version of Gene Ontology terms for functional annotation across organisms.
What GO is
LiteFold/GO provides a tabular representation of the Gene Ontology (GO), a controlled vocabulary and knowledgebase that defines gene product functions, biological processes, and cellular components across living organisms. Produced by LiteFold, the resource transforms hierarchical Open Biomedical Ontologies (OBO) data into structured Parquet tables.
Functional genomics studies often require programmatic access to ontology definitions, hierarchies, and synonyms without parsing complex flat files. By parsing each ontology stanza into row-level tabular data, this dataset simplifies computational workflows in bioinformatics and machine learning.
The dataset reflects the GO releases/2026-03-25 release, capturing 48,291 total terms across three namespaces: biological process, molecular function, and cellular component. Both active terms and obsolete entries are recorded alongside graph parentage and cross-reference metadata.
How it was made
The dataset is derived directly from the canonical go.obo and go-basic.obo files. Each row represents a single [Term] stanza, with repeated OBO attributes formatted as list columns. Preserved fields include term definitions, synonym mappings with their associated scopes (such as EXACT, BROAD, NARROW, or RELATED), and cross-references.
Graph relationships are captured through direct taxonomic is_a identifiers alongside predicates such as part_of and regulates. These links are combined into unified parent identifier lists. Relationship typedef metadata from go.obo is separately catalogued as a dedicated metadata table. Deterministic train and test splits are assigned via a hash bucket: sha256(go_id) modulo 10.
Key facts
- Publisher
- LiteFold
- GO Release
- releases/2026-03-25
- Total Terms
- 48,291
- Active Terms
- 38,560
- Obsolete Terms
- 9,731
- Format
- Parquet
- Biological Process Terms
- 30,857
- Molecular Function Terms
- 12,839
- Cellular Component Terms
- 4,595
- Train Split Size
- 43,522 rows
- Test Split Size
- 4,769 rows
What you can use it for
- Ontology Graph Traversal
- Construct directed acyclic graphs of functional terms using parsed is_a relationships and predicate relationship targets.
- Genomic Text Mining and Entity Linking
- Match biological entities in scientific literature against standardised term names, alternative identifiers, and scoped synonyms.
- Functional Term Filtering
- Extract active subsets of specific namespaces, such as biological processes, to annotate gene expression or proteomics datasets.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from LiteFold/GO at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Converts unstructured OBO stanzas into typed Parquet columns and lists without requiring specialised flat-file parsers.
- Retains full ontological linkage metadata, including parent relationships, obsolete term replacements, and typedef definitions.
- Provides deterministic, reproducible data splits based on SHA-256 hashes of GO identifiers.
Watch out for
- Contains 9,731 obsolete terms that require explicit filtering via the is_obsolete flag to avoid deprecated functional concepts.
- Covers only ontology vocabulary and structure, omitting direct gene-to-term annotation associations.
Questions
Can I look inside GO without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is GO under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
What namespaces are present in the dataset?
The terms are divided across biological_process (30,857 rows), molecular_function (12,839 rows), and cellular_component (4,595 rows).
How are obsolete terms handled?
Obsolete terms are retained in the dataset and marked with the is_obsolete boolean flag. They contain fields for suggested replacement IDs (replaced_by and consider) where available.
Are the original OBO files retained?
Yes, the repository retains the underlying go.obo and go-basic.obo files alongside the Parquet tables.
How is the train/test split determined?
The split is assigned deterministically by calculating sha256(go_id) % 10. Bucket 0 is allocated to the test set, while buckets 1 through 9 form the training set.
How to cite it
The Gene Ontology Consortium, The Gene Ontology knowledgebase in 2023, Genetics, 224(1), 2023, doi:10.1093/genetics/iyad031
Written from GO's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.