Cannot replicate prior results3

Question

Cannot replicate prior results3

karafecho opened this issue 4 years ago · 8 comments

karafecho commented 4 years ago

This issue is essentially the same as #552 and #553 except for isopropranolol - gene - allergic disease.

karafecho commented 4 years ago

Thanks!

Answer 1 · 2021-04-16T15:59:50.000Z

Here is a link to the answers.

Answer 2 · 2021-04-16T16:07:37.000Z

Note that the answer set is missing the gene DDX39B, which was returned with the original query.

Answer 3 · 2021-04-19T14:22:01.000Z

Do you happen to have any information on the original result? maybe the link to the original question? Trying to track down DDX39B, and it would be helpful to know where the edges originally came from.

Answer 4 · 2021-04-19T14:45:00.000Z

OK, so I think I see what's going on here. It has to do with a change in how we decided to handle subclasses. We used to add subclass-inferred edges into the database directly. So, DDX39B has an edge from biolink that it "causes" atopic eczema (MONDO:0004980), which is a subclass of Allergic Disease (MONDO:0005271) (see https://www.ebi.ac.uk/ols/ontologies/mondo/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FMONDO_0004980).

We used to infer a DDX39B-[causes]->AllergicDisease edge and add it to the graph, but we decided that this was messy and better handled elsewhere. We do keep the subclass edges in the graph, as you can see from this query: https://robokop.renci.org/a/5986af83-34d5-4f4c-bce7-78930af62102_60a5cb79-9f78-43ac-966b-3e5e99995d1e/ so in principle, a question that added another disease node between the gene and allergic disease, with a subclass edge to allergic disease should reproduce the initial result.

Unfortunately, it looks like there is also some kind of load or versioning problem that did not create a subclass of edge between atopic eczema and allergic disease. We could rectify this perhaps by reloading the MONDO subclass of edges, in which case the query outlined above would work.

Going forward, when the new robokop is talking to a RKG being served from automat/plater, the subclass of edges will be followed automatically in accordance with the agreed upon KP behavior, but this is not yet implemented.

Answer 5 · 2021-04-19T15:45:32.000Z

Thanks for looking into this. To clarify, is the plan to reload the MONDO subclass of edges? Or, should I simply explain the issue in the rebuttal letter?

Answer 6 · 2021-04-19T18:05:35.000Z

Looking into it - it's a bit complicated because the robokop graph is old and so it requires some backporting.

Answer 7 · 2021-04-20T17:47:01.000Z

This has now been updated. Here's a link to a new version of the question that includes explicit subclassing: https://robokop.renci.org/q/8d5556c4-2826-491b-99a0-941086840239/. It includes DDX39B, but also looks like it may add in another gene as well.