Query will not run in ROBOKOP but is executable in ROBOKOP KG

[karafecho]

I attempted to use ROBOKOP to answer the MCAT question below.

MCAT Question Steatorrhea is the presence of increased fat in feces. Which organ is least likely to be the cause of a patient’s steatorrhea?

Correct: Stomach

Incorrect: Pancreas; Small intestine; Liver

Initially, this question was problematic due to a lack of identifiers for stomach, pancreas, and liver; since fixed on 2/13/2020. However, the query still does not run, even when I do not specify "anatomical entity".

[cbizon]

A little more information. Here's a robokop query that returns no answers:

https://robokop.renci.org/q/3692d7a6-83c1-4c5b-8641-66d718a55f8a/

But, if I run the query in robokopdb2 (I think the live one?) then I get answers:

match (p:phenotypic_feature {id:"HP:0002570"})<-[:has_phenotype]-(d:disease)-[:located_in]->(a:anatomical_entity) return * limit 10

[patrickkwang]

Perhaps some strange synonymization is happening here. This is the query graph:

  {
    "nodes": [
      {
        "id": "n0",
        "type": "disease_or_phenotypic_feature",
        "set": false,
        "name": "Steatorrhea",
        "curie": [
          "HP:0002570"
        ]
      },
      {
        "id": "n1",
        "type": "disease",
        "set": false
      },
      {
        "id": "n2",
        "type": "anatomical_entity",
        "set": false
      }
    ],
    "edges": [
      {
        "type": [
          "has_phenotype"
        ],
        "id": "e0",
        "source_id": "n1",
        "target_id": "n0"
      },
      {
        "type": [
          "located_in"
        ],
        "id": "e1",
        "source_id": "n1",
        "target_id": "n2"
      }
    ]
  }

and this is the derived Cypher query:

MATCH (n1:disease {})-[e0]->(n0:disease_or_phenotypic_feature {})
USING INDEX n0:disease_or_phenotypic_feature(id)
WHERE (n0.id = 'MONDO:0001075') AND (type(e0) = "has_phenotype")
MATCH (n1)-[e1]->(n2:anatomical_entity {})
WHERE (type(e1) = "located_in")
WITH n0.id as n0, n1.id as n1, n2.id as n2, collect(distinct e0.id) as e0, collect(distinct e1.id) as e1
RETURN {n0:n0, n1:n1, n2:n2} as nodes, {e0:e0, e1:e1} as edges

[cbizon]

Hmm. @YaphetKG are we by any chance using the production redis server for the new robokop mini- crawls?

[YaphetKG]

we are using another instance for those,

[karafecho]

Thanks, all.

Note that I kind of need this issue resolved sooner rather than later so that I can complete my second MCAT 'experiment' by the end of the month and include the results on a poster that I am preparing for the AMIA 2020 Informatics Summit (you all are authors, btw). Poster will then be turned into a journal article. (Think: incentive.) :-)

However, I don't want to distract from other activities, so if the issue will require too much time to resolve, then I'll develop Plan B.

[karafecho]

@patrickkwang @YaphetKG @cbizon : Any status updates on this issue?

[cbizon]

It's going to take some time to fix. Can you try using the MONDO term MONDO:0001075 for steatorrhea instead?

[karafecho]

I had already tried the MONDO ID.

I've given up on the original query, at least for the time being. I am probably going to use this answer list. Thoughts?

[karafecho]

But check out what happens when I add the predicate "gene_to_expression_site_association".

[cbizon]

I had already tried the MONDO ID.

And what happened in that case? Still no results?

[karafecho]

Sorry, yes, the MONDO ID yielded no results. I tried a few query permutations with both the MONDO ID and the HP ID---nothing ran.

But the newly structured query works: MONDO 0001075 - gene - anatomical entity.