Streamline dipper and Unify OBAN association modeling

[cmungall]

(this description may be iteratively refined)

Currently we follow the OBAN association model (essentially RDF reification) extending it with source-specific information. Cypher queries translate chains of associations into denormalized triples for loading into Solr. The overall process is not very transparent.

We have configuration and tests distributed across

• RO (property chains, subprops), e.g. transitive properties
• GENO (ditto)
• dipper (cmaps) e.g. kegg
• monarch-cypher-queries/src/cypher/$SUBCAT-$OBJCAT.yaml e.g. gene-phenotype
• monarch-app/conf/golr-views/single-tab/$SUBCAT-$OBJCAT.yaml e.g. gene-phenotype
• biolink routes
• biolink tests

Proposal

Preliminary definitions:

• category-pair: a tuple of (subject_category, object_category)

We will have a master location (in dipper repo?) with one YAML file per category pair. This YAML file will compile down to

• cypher query, which will be a UNION of subject-object paths
• a golr-view spec file in YAML
• one or more behave tests for biolink-api - ideally each path will be tested at least once
• a markdown documentation file (including the behave test - see for example biolink examples )
• one or more example dot/png graphs generated from the test case (the evidence graph)
• validation against RO?
• flaskrest code (but use for manual sync for now)

[TomConlin]

I approve of this message