Write a piece "Towards a universal validator of identifiers used in research"

[Daniel-Mietchen]

Gist:

• identifier systems exist in many parts of society, for physical and digital entities, for processes, locations, natural or human-made structures and so on
• many of these are used in research publications
• some are wrong, in one or more of various ways
• usage of any such identifiers rests upon certain assumptions, including a reliable association between the entity to be identified and its respective identifier
• sometimes, that association is missing or otherwise incorrect
• in principle, we could have a system with the following characteristics
  • a registry of identifier schemes used in scholarly publications
  • for each identifier scheme, it would contain information about
    • its scope (e.g. some statement about the classes of entities for which the scheme is intended)
    • permissible identifiers (e.g. as given via a regular expression)
    • a mechanism to validate whether
    • a given identifier
      • is valid
      • meets a pre-specified set of other criteria (e.g. uniqueness within the scheme)
    • can be resolved to a resource providing mechanisms to assess the quality of the match between the identifier and the entity it is meant to identify

Examples I have come across:

• DOIs (of works and cited works)
• ORCIDs (invalid ORCIDs, valid ORCIDs used in invalid contexts, multiple uses of valid ORCIDs )
• post codes
• accession numbers in places like GenBank, PDB, Uniprot
• ethics approval numbers
• cell line identifiers
• specimen codes in clinical or museum settings

[Daniel-Mietchen]

I have started a collection of cases where various identifiers have been used incorrectly: https://github.com/br2s/bug-reports-to-science/blob/master/problem-categories/identifiers/generic.md .

I have also begun to look into use cases where this might matter, e.g. A botanical demonstration of the potential of linking data using unique identifiers for people

[Daniel-Mietchen]

Some further examples from Functionathon: a manual data mining workflow to generate functional hypotheses for uncharacterized human proteins and its application by undergraduate students:

The tutors identified a mistake in mouse genome databases (fam90a1b wrongly annotated as a CXorf58 ortholog) and one in the literature (TMEM232 wrongly described as a tetraspan protein) that would have been difficult for the students to spot. While MGI was notified and the mistake will be corrected, mistakes in the literature are generally not corrected.