Bioregistry

The bioregistry Python package provides programmatic access to the Bioregistry meta-registry. It is designed to help developers perform common life science nomenclature tasks, including:

• Metadata lookup: Retrieve information about databases, ontologies, and other resources.
• CURIE expansion: Convert Compact URI Identifiers (CURIEs) into full URIs.
• URI contraction: Convert full URIs into CURIEs.
• Other registry operations: Accessing the integrative, community-driven meta-registry of life science resources.

The bioregistry project includes benchmarks to measure the performance of its core parsing and validation functions. There are three primary benchmark suites:

1. uri_parsing: Evaluates the performance of the bioregistry.parse_iri function.
2. curie_parsing: Evaluates the performance of the bioregistry.parse_curie function.
3. curie_validation: Evaluates the performance of the bioregistry.is_valid_curie function.

The exports/analyses directory contains results from various automated analyses used for curation and quality control. Developers can inspect these files to understand the current state of the registry's data quality and curation status.

Available analyses include:

• paper_ranking: Uses TF-IDF and a binary classifier to rank new PubMed papers for potential curation.
• mapping_checking: Uses embedding-based similarity checking to verify mapped prefixes for curation.

The Bioregistry SSSOM export provides ontology mappings in the Simple Standard for Sharing Ontology Mappings (SSSOM) format. The data is distributed in two primary files:

• Mappings (TSV): Contains both predicted and curated mappings, located in bioregistry.sssom.tsv.
• Metadata (YAML): Contains the CURIE Map, credits, and other metadata, located in bioregistry.sssom.yml.

| Facet        | Description                      | Link                                             |
| ------------ | -------------------------------- | ------------------------------------------------ |
| TSV          | Mappings (predicted and curated) | [`bioregistry.sssom.tsv`](bioregistry.sssom.tsv) |
| Metadata YML | CURIE Map, credits, etc.         | [`bioregistry.sssom.yml`](bioregistry.sssom.yml) |

The CURIE Validation Benchmark is used to verify the correctness of the bioregistry.is_valid_curie function. It evaluates whether the function accurately identifies valid and standard CURIEs (Compact Uniform Resource Identifiers) according to Bioregistry standards.

Dataset Structure

The benchmark uses a dataset (data.tsv) containing the following columns:

• prefix: A canonical Bioregistry prefix.
• identifier: A local unique identifier within that prefix's semantic space.
• label: A boolean (true or false) indicating if the prefix-identifier combination is a valid and standard CURIE according to Bioregistry.
• curie: The formatted CURIE string for the pair.

The CURIE Parsing Benchmark is used to evaluate the performance of the bioregistry.parse_curie function. It measures how many CURIEs (Compact URI) the parser can process per second using a specific dataset of prefixes, identifiers, and synonyms.

Performance typically averages around 4,000 CURIEs per second.

The Bioregistry provides aggregated data regarding the countries of the institutions responsible for the resources indexed in the registry.

Important Note: This information is not a requirement for indexing, so it is missing for a large number of entries. The presence of this data is a result of community curation rather than a mandatory field for all resources.

Providers should be curated when you want to expose specialized endpoints or handle specific URI patterns. Use providers for:

1. Additional endpoints: Accessing different types of information such as image resolution, JSON/XML/structured data artifacts, or other specialized views.
2. Data-specific URI patterns: Handling URI patterns that appear frequently in datasets but are not necessarily intended for primary resolution.

Note on Subsets: It is acceptable to curate providers that only resolve a subset of the semantic space (e.g., a provider that only resolves UniProt identifiers for antigens). If you do this, ensure the description field clearly states the limitations.

The Bioregistry's metaregistry contains alignments from external registries (e.g., Identifiers.org, OBO Foundry, FAIRsharing). You can contribute by reviewing these alignments and determining if a record should be:

1. Mapped to an existing Bioregistry prefix.
2. Imported as a new external prefix.
3. Blacklisted if it is irrelevant.

Alignment data for external registries can be found in the exports/alignment*.tsv files in the repository.