Align and connect 'supporting document' and 'publications' edge properties

[mbrush]

At present, the most common Biolink edge property used in Attributes to hold supporting publications is publications.

But Biolink also provides a supporting document edge property that would cover non-publication source documents (a request from TMKP where not all mined text is from publications).

At present these edge properties are not hierarchically related in the Biolink Model.

Questions:

1. Do we want to keep both?
2. If so, we should make publications a child of supporting document. And name these properties consistently (e.g. change 'publications' -> 'supporting publications')
3. If not, which one to keep?
4. If we decide to only keep 'supporting document' do we need a way to indicate that the document is a publication? a. e.g. Attribute.value_type field = biolink:Publication b. e.g. a nested Attribute object keyed on the biolink:type property, with value = biolink:Publication

[mbrush]

Proposal:

• Keep both, but rename publications -> supporting publications
• Move supporting publications under supporting documents
• Update definitions to make them more consistent with each other and their new hierarchical relationship.
  • Proposed supporting documents definition: A referencable document that reports the statement expressed in an Association, or provides information used as evidence supporting this statement.
  • Proposed supporting publications definition: A publication that reports the statement expressed in an Association, or provides information used as evidence supporting this statement.

Another issue here is to decide if the names of these properties should be plural or singular . . .

[sierra-moxon]

@mbrush - is it necessary to change 'publications' to 'supporting publications'? (totally doable, I just know this is a very widely used parameter and would require a lot of downstream changes. Alternatively, I could add supporting publications as an alias of publications).

[sierra-moxon]

also should (somewhat pedantically) note that supporting document is currently in the model, and is not multivalued. Would the second part of this proposal make supporting documents multivalued?

[sierra-moxon]

should 'supporting text' also become 'supporting texts' and be a child of 'supporting documents'?

[sierra-moxon]

should 'supporting data set' also be a child of 'supporting documents' ?

[mbrush]

My thoughts on your questions @sierra-moxon:

• not totally necessary to change publications to supporting publications, but IMO it is preferred (it is a more descriptive/informative name, and provides consistency with supporting documents)
• supporting documents was created at the request of TMKP, and they did not need it to be multivalued. But I think it should be, so support other use cases for this property.
• supporting text should not be renamed, and is not a subproperty of supporting document (it refers to a single span of text, not an entire document)
• supporting data set should not be moved under supporting document either, as IMO we should not consider a dataset to be a type of document. (This is consistent with how they are modeled in OBO ontologies like IAO)

[mbrush]

Re-opening this issue as additional questions regarding the representation of supporting documents have surfaced, and may warrant reconsideration of decisions made above.

Current Biolink representation:

### CLASSES ###

  publication:
    is_a: information content entity
    description: >-
      Any published piece of information. Can refer to a whole publication,
      its encompassing publication (i.e. journal or book) or to a part of a
      publication, if of significant knowledge scope (e.g. a figure, figure
      legend, or section highlighted by NLP). The scope is intended to be
      general and include information published on the web, as well as printed
      materials, either directly or in one of the Publication Biolink
      category subclasses.
    slots:
      - authors
      - pages
      - summary
      - keywords
      - mesh terms
      - xref
      . . . 

### ASSOCIATION SLOTS  ###

 supporting documents:
    is_a: association slot
    description: >-
      One or more referencable documents that report the statement expressed in an Association, or provide 
      information used as evidence supporting this statement.
    range: uriorcurie
    multivalued: true
    examples:
      - value: PMID:12345678

  publications:
    aliases: ["supporting publications"]
    singular_name: publication
    description: >-
      One or more publications that report the statement expressed in an Association, or provide information used as 
      evidence supporting this statement.
    is_a: supporting documents
    multivalued: true
    range: publication

Questions:

1. Confirm we want two slots to link to supporting docs/pubs.

2. Revisit idea or renaming to be more consistent.

3. Broaden supporting documents and/or publications definition to include the idea of documents supporting things besides Statements (e.g. StudyResults), and indicating that provenance information qualifies as supporting info:

Documents that provide information supporting the creation of a Statement or other Information Entity. Most often this slot is used to reference documents that report the Statement expressed in an Association, or provide evidence or provenance information supporting this Statement. Another common use is to reference documents that supported the creation of a Study Result object.

An issue here is that the publications property currently an association slot, as its initial scope is to annotate Edges/Statements. But for linking other types of informational entities to pubs that support them (e.g. Study Results . . . .which are named things, not associations), an association slot is not appropriate (we would need a node property or predicate).

4. Establish conventions of use for these properties in TRAPI messages (see tickets here and here)

[mbrush]

Voting on decisions that address 1-4 above is happening in Discussion #1234.

[mbrush]

Closing, based on outcomes of voting in #1234 and final decisions documented in the supporting publications specification here.