Add Node Normalizer CURIE synonyms to Node attributes

[gprice1129]

Problem
There was an expectation by the UI team that the CURIE synonyms for a given node would be provided by the reasoners. This doesn't appear to be happening in all cases. Since the ARS already calls the Node Normalizer for a result, it has this information at hand and should be able to easily tack on the CURIE synonyms to the result.

Proposed Solution
Add the CURIE synonyms in a biolink:xref attribute. There is no requirement from the UI for biolink:xref attributes to be unique meaning either of the following would be fine:

1. The ARS merges it with an existing biolink:xref attribute
2. The ARS blindly creates its own biolink:xref attribute

Example provided below for reference:

"MONDO:0001302": {
  ...
  "attributes": [
    ...
    {
      "attribute_type_id": "biolink:xref",
      "value": [
        "MONDO:0001302",
        "DOID:11516",
        "UMLS:C0152105",
        "UMLS:C0694541",
        "MEDDRA:10020801",
        "MEDDRA:10020823",
        "MEDDRA:10020824",
        "MEDDRA:10020827",
        "MEDDRA:10045916",
        "NCIT:C157879",
        "NCIT:C4907",
        "SNOMEDCT:275516004",
        "SNOMEDCT:64715009",
        "medgen:508889",
        "ICD10:I11",
        "ICD9:402"
      ]
    },
    ...
   ],
   ...
}

[github-actions]

Thanks! Feedback is very much appreciated!

[cbizon]

Could we use something other than xref? @sierra-moxon is there a better biolink attribute for identifier equivalence?

[ShervinAbd92]

@cbizon as a matter of fact, ARS currently adds the synonyms to the attributes under biolink:same_as and biolink:xref is denoted to the original_id. i was planning to put everything under xref but if you guys prefer the current naming convention, i can leave it.

   "MONDO:0035838": {
              "attributes": [
                  {
                      "attribute_type_id": "biolink:xref",
                      "original_attribute_name": "original_id",
                      "value": [
                          "UMLS:C5197665"
                      ],
                      "value_type_id": "metatype:NodeIdentifier",
                      "attribute_source": null,
                      "value_url": null,
                      "description": null
                  },
                  {
                      "attribute_type_id": "biolink:same_as",
                      "original_attribute_name": "equivalent_identifiers",
                      "value": [
                          "MONDO:0035838",
                          "orphanet:570431",
                          "UMLS:C5197665",
                          "medgen:1680858"
                      ],
                      "value_type_id": "metatype:NodeIdentifier",
                      "attribute_source": null,
                      "value_url": null,
                      "description": null
                  }
              ],
              "categories": [
                  "biolink:Disease"
              ],
              "name": "idiopathic multicentric Castleman disease"
          }

[gprice1129]

If this is what biolink:same_as is supposed to denote than I suggest we just leave it like that.

[ShervinAbd92]

sounds good @gprice1129

[gprice1129]

@ShervinAbd92 from what you've said it sounds like this is already complete, but there are cases on Production where there is no biolink:same_as attribute even though NN has many synonyms.

[ShervinAbd92]

@gprice1129 Thats interesting. Can you share an example that you are seeing this behavior in Prod? so i can test it out the query on my end

[gprice1129]

Here is one example: f6888343-18ad-4989-aae0-47cb090d8ddc. Look at the node CHEBI:6888 (Methylprednisolone)

[ShervinAbd92]

@gprice1129 oki so looking into this query, i see Improving, ARAX, BTE and Molecular KP have returned Methylprednisolone in their results. looking deeper into their kg nodes i can see that for improving and molecular kp, ARS node norm added biolink:xref (original_id) and biolink:same_as (equivalent_identifer) but for BTE and ARAX the returned normalized nodes had already came back to us with attributes_type_id: synonyms (curie strings) and biolink:xref (which seems to include also the same synonyms we have added to improving and molecular under 'same_as' attribute! now i am thinking in order to have uniformity between these normalized ARAs and their attribute type_id naming convention, your initial idea on putting the synonyms under xref can be a good idea.

[gprice1129]

@ShervinAbd92 in the merged results for CHEBI:6888 I don't see any attributes other than the ones from the annotator.

[ShervinAbd92]

@gprice1129 yeah i can see that probably our merge process is not handling the attributes properly. give us some time to look into it.

[MarkDWilliams]

@gprice1129 apologies for the delay, but this should now be fixed in CI.