8 Do not re assign or delete identifiers
Permalink URI: https://w3id.org/id-rules/8
Identifiers generated and publicly advertised must never be reassigned to a different record or deleted. If you issue identifiers, consider their full lifecycle: there is a fundamental difference between identifiers which point to experimental datasets (GenBank/ENA/DDBJ, PRIDE, etc.) and identifiers which point to a current understanding of a biological concept (Ensembl Gene, UniProt record, etc.).
While experimental records are less likely to change, concept descriptions may evolve rapidly; even the nature and number of the relevant metadata fields change over time. Moreover, the very notion of identity is often strongly impacted by relationships (e.g., between concepts or processes).
Extensive changes cannot be captured with numerical suffixing alone. For instance, taxonomists may split or merge species, pathologists may split or merge diseases, or hypothesized entities may be proven not to exist (e.g. vaccine-induced autism). Global initiatives are actively exploring identifier strategies for such use cases. In the meantime, consider these recommendations:
Table 3:
Recommended handling | Example |
---|---|
Obsoletion : If an entry has been removed or deprecated, the original identifier must still resolve to a ‘tombstone page’. Reasons for obsolescence should be indicated. If the obsoleted ID is replaced by another ID, the replacement must be present and also described as automatic or suggested, preferably using the ontology properties iao:replaced_by and obo:consider, respectively. | uniprot:A0AV18 |
The obsoleted ID must never be reassigned to another entity. A list of obsoleted IDs should be maintained. | Obsoleted Uniprot identifiers |
Merging: When two or more identifiers are merged, a new recipient identifier should be designated as the primary (citable) one and should contain information about the legacy identifiers it encompasses. Any legacy identifiers should continue to resolve via redirection to the primary identifier. | uniprot:Q57339 and uniprot:O08022 merged into uniprot:Q00626 |
Splitting: If an identifier is split (demerged) into two or more new ones, new identifiers should be assigned to all the new entries. The legacy identifier must be obsoleted, must resolve, and should provide a warning and pointers to the new ones as per above. | uniprot:P29358 split into uniprot:P68250 and uniprot:P68251. |
This section was edited from the original to simplify the examples. Uniprot URIs updated to use purl.uniprot.org
10 Simple rules for design, provision, and reuse of identifiers for web-based life science data by Julie McMurry, et al. is licensed under a Creative Commons Attribution 4.0 International License.
Based on a work at http://dx.doi.org/10.5281/zenodo.31765.
Adaptation from the original preprint to wiki format include typography modifications, changing citations to updated hyperlinks, splitting sections/paragraphs and moving tables/boxes. Modifications by Stian Soiland-Reyes, eScience lab, University of Manchester.