Ontology · LiteFold
GOA
Evidence-coded Gene Ontology annotations for UniProtKB proteins, RNAs, and protein complexes.
What GOA is
Gene Ontology Annotation (GOA) provides evidence-coded annotations associating UniProt Knowledgebase (UniProtKB) proteins, RNAs, and protein complexes with terms from the Gene Ontology (GO). These annotations link gene products to standardized terms describing biological processes, molecular functions, and cellular components.
This distribution packages the full GOA UniProt source files alongside a tabular sample table in Apache Parquet format. The Parquet table contains 100,000 parsed annotation rows designed for schema inspection, query prototyping, and quick exploratory analysis.
For exhaustive functional genomics analyses, the dataset provides direct access to the complete compressed Gene Association Format (GAF) and Gene Product Association Data (GPA) source files comprising the full UniProt annotation repository.
How it was made
The dataset contains two primary compressed archives: goa_uniprot_all.gaf.gz (15,387,303,487 bytes) and goa_uniprot_all.gpa.gz (9,462,421,263 bytes), alongside a source-file manifest in metadata/source_files.parquet.
The default Parquet table parses the initial 50,000 rows from each raw format into structured columns capturing identifiers, evidence codes, reference links, qualifiers, and taxonomic information. Each record is indexed with a deterministic SHA-256 identifier, with a modular hash (sha256(annotation_id) % 10) allocating bucket 0 to the test split (10,042 rows) and buckets 1 to 9 to the train split (89,958 rows).
Key facts
- Publisher
- LiteFold
- Sample rows
- 100,000
- Train split size
- 89,958 rows
- Test split size
- 10,042 rows
- GAF source size
- 15,387,303,487 bytes
- GPA source size
- 9,462,421,263 bytes
- File formats
- Parquet, GAF (gzip-compressed), GPA (gzip-compressed)
- Modalities
- Tabular, text
What you can use it for
- Schema inspection and pipeline prototyping
- Inspect column schemas, evidence codes, and qualifier structures using the Parquet sample without downloading tens of gigabytes of raw files.
- Protein function lookup
- Retrieve Gene Ontology terms and associated evidence codes for specific UniProt identifiers, RNAs, or complexes.
- Ontology aspect filtering
- Filter parsed records by specific Gene Ontology domains, such as molecular function (F), biological process (P), or cellular component (C).
- Proteome-wide functional genomics
- Download and process the complete GAF or GPA compressed archives for comprehensive biological entity annotation.
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/GOA at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Supplies standardized, evidence-coded Gene Ontology annotations for proteins, RNAs, and complexes across UniProtKB.
- Provides a manageable Parquet sample table for quick local exploration alongside complete, multi-gigabyte source archives.
- Employs deterministic hashing to create fixed, reproducible train and test evaluation splits.
Watch out for
- The default Parquet table covers only 100,000 sampled rows, requiring users to download the multi-gigabyte raw files for complete coverage.
- Certain fields, including aspect, object name, and object synonyms, are only available for GAF records and are missing from GPA records.
Questions
Can I look inside GOA 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 GOA under?
Its authors declare other. Read the terms on the original page before reusing the data.
What is the difference between the Parquet table and the compressed files?
The Parquet table contains the first 50,000 parsed rows from both the GAF and GPA files for quick querying and schema preview. The full compressed GAF and GPA archives contain the entire collection of GOA UniProt annotations.
How are the training and test splits partitioned?
The split is generated deterministically by applying sha256(annotation_id) % 10. Bucket 0 forms the test split (10,042 rows), while buckets 1 through 9 form the train split (89,958 rows).
What biological entities are covered in this dataset?
The dataset contains annotations for UniProtKB proteins, RNAs, and multi-protein complexes.
What information does the aspect column contain?
In GAF records, the aspect column specifies the Gene Ontology domain: F for molecular function, P for biological process, or C for cellular component. It is not populated for GPA rows.
How to cite it
Huntley, Rachael P., Tony Sawford, Prudence Mutowo-Meullenet, Aleksandra Shypitsyna, Carlos Bonilla, Maria Jesus Martin, and Claire O'Donovan. 'The GOA database: Gene Ontology annotation updates for 2015.' Nucleic Acids Research 43, no. D1 (2015): D1057–D1063. doi:10.1093/nar/gku1113.
Written from GOA'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.