Action list from January 2024

[RichardBruskiewich]

1. • Identify or create a more specific predicate for the SemSimian output knowledge edges
     • For now use biolink:similar_to
     • Review better options for this, e.g. biolink:manifestation_of or ???
2. • Review and implement the (Vlado et al) UUID identification strategy for description of the UUID generation of subject or object nodes of an edge/association based on a set of terms members of a phenotypic profile.
     • Review interpretation and implementation of the is_set flag in the context of the multi-CURIE use case.
     • Generate and apply UUID nodes for knowledge graph and result (phenotypic) profile subsets of terms
3. • Specify a complete sample TRAPI Response JSON to guide further complete TRAPI /query implementation _(Matt, Sierra)
     • Add the 'auxiliary_graphs' JSON block to the TRAPI response
     • Build and return TRAPI auxiliary graphs containing detailed similarity assertions with scores from SemSimian
     • Review the need to somehow report (unduplicated?) the symmetrical subject_best_matches in addition to the object_bestmatches already returned.
4. • Use infores:semsimian-kp as the primary_knowledge_source (How do we account for the original knowledge assertions, e.g. from HPO or MONDO or ???)
5. • By default, only return “the top 5” most similar pairwise comparisons to limit the size of the auxiliary graph block, however, incorporate some kind of TRAPI QueryGraph parameter to signal expansion of results to a larger number of hits.
6. • Convert vanilla Docker to Docker Compose management of the containers.
7. • Fill out the template for the January Relay presentation
8. • Fix server Open Telemetry code - the latest OT code modules appear to have removed the trace.get_tracer_provider().add_span_processor method
9. • If desired, implement the 'common' API endpoints - i.e. get_node() and get_single_hops()
10. • Review whether or not a new specific "multi-curie' workflow operation needs to be defined for this use case
11. • Decide on official name of this system including KP versus ARA nature
12. • Review meeting minutes for any implementation or feedback loose ends.
13. • Review latest Plater code for any additional changes worth incorporating

[RichardBruskiewich]

@mbrush and @sierra-moxon, a super quick scan of the Biolink Model dredged up the following predicate:

  manifestation of:
    is_a: related to at instance level
    description: >-
      that part of a phenomenon which is directly observable or
      visibly expressed, or which gives evidence to the underlying
      process; used in SemMedDB for linking things like dysfunctions
      and processes to some disease or syndrome
    range: disease

which almost feels like an appropriate predicate, even though it is not taken as 'similarity' per say.