Retrieve data without a relation/predicate

[kshefchek]

In the current implementation, data without a relation is filtered out. Would it be possible to include this data and allow it to be empty or null?
https://github.com/NCATS-Tangerine/biolink-beacon/blob/de04cf1/main.py#L121

Also, phenotype should be lower case when accessing the category in solr and/or biolink.
https://github.com/NCATS-Tangerine/biolink-beacon/blob/de04cf1/main.py#L335

[mbrush]

@kshefchek I think this speaks to a question I had about whether 'inferred' triples (associations inferred via cypher queries) would show up in the beacon aggregator/tk.bio. I assume that these are the types of associations/triples you have in mind?

Important that these show up as these are a large proportion of the data served by Biolink. Also speaks to the issue here about the utility of named relations describing all inferred associations.

[cmungall]

What would not have a predicate?

An inferred axiom still has a predicate.

[mbrush]

@cmungall I realize that my description above of "inferred triples" that lack a predicate makes no sense (as without a predicate they are not triples!). To clarify, I'm thinking of 'inferred' associations generated via cypher queries and loaded into solr, which then get returned by biolink. For example, a gene-disease association 'inferred' by the cypher query here:

MATCH path=(subject:gene)<-[geno:GENO:0000418|GENO:0000639!]-(feature)-[:RO:0002200|RO:0002610|RO:0002326|RO:0003302!]->(object:disease)  
RETURN path, subject, object, 'gene' AS subject_category, 'disease' AS object_category, 'inferred' as qualifier

This association has no predicate assigned - such that the 'relation' object in biolink calls is null, e.g. see https://api.monarchinitiative.org/api/bioentity/gene/NCBIGene%3A4750/diseases/?rows=20&fetch_objects=true

I think these are the types of associations that @kshefchek is referring to, which the k-beacon aggregator is filtered out.

[kshefchek]

In retrospect, we can leave this constraint in, as the next monarch release should have near full coverage of relations linking subject object pairs. However, we still need to correct the Phenotype -> phenotype issue.