Closed karafecho closed 4 months ago
This was due to content differences between the graphs hosted at robokopkg.renci.org (the graph exemplar queried) and robokopkg on Automat (the graph qgraph queried).
I think this needs to be reconciled. We should not have two different versions of ROBOKOP KG.
Moreover, we've branded robokopkg.renci.org as the ROBOKOP KG, so that endpoint should be ground truth.
Decision on 04/10/2023: robokopkg.renci.org and http://automat.renci.org/#/robokopkg will point to duplicate copies of a new ROBOKOP KG that will include the existing Automat ROBOKOP KG plus Text Miner KG (but not CORD19 KG).
I think that this has been implemented.
Please see my last two posts in RobokopU24/Use-Cases#10. I haven't been able to replicate Jon-Michael's ExEmplar results using ROBOKOP qgraph. Note that the screenshots I captured include only a few of the queries I tried. In addition to testing a variety of predicates and nodes (e.g., herpes zoster vs shingles), I tested several different chemical entity selections (i.e., from the duplicates in the drop-down list).
I'm not expecting results for all five of the input chemical entities, but I wouldn't think that would break a query.
Subsequent troubleshooting with Jon-Michael revealed the following.
This ExEmplar query hits the ROBOKOP KG Neo4j db directly and returns results:
This ROBOKOP qgraph query does not return results:
Same query as the ExEmplar one, but run in the Neo4j browser, with results returned:
Same query as the ExEmplar one, but posted to ROBOKOP KG on Automat, with no results returned:
I'm thinking this might be a Cypher to ROBOKOP qgraph / TRAPI translation issue?