handling ID-prefixes in BTE in a consistent, straightforward way

[colleenXu]

BTE currently:

• in most cases, it represents entity IDs (like a specific disease, gene, etc) in curie format (prefix:ID). Specifically, it uses the Translator/biolink-model capitalization and spelling for the prefix + ID
• but in a few cases, including sub-query generation, BTE will remove the prefix for specific ID-namespaces while keeping it for others
  • not sure if we've figured out all the areas where this happens, or all the areas where the ID is required to be a certain format in order for things to function....
  • but sub-query generation + api-response-transform are two known important areas, because the APIs we're querying may not use ID-prefixes or the capitalization/spelling that Translator does...

This can easily lead to bugs and is confusing. With the current big code update, it's not clear if there are some bugs / unexpected behavior with ID-prefixes....which prompted the discussion + writing this issue.

---

2023-03-22 discussions in group meeting and afterwards:

• getting the shared understanding written above
• I propose that in the case of sub-query generation (specifically the format of string values in queryInputs), BTE should remove the prefix from all ID-namespaces (rather than keeping prefixes for some)
  • It'll be easier to adjust the x-bte annotation because I can then search repos for the prefixes in the linked list and adjust only those operations
• @tokebe said we'll still need to review what's going on with ID-prefix handling before making decisions / code-changes.

Seems like a design decision between @tokebe and me, although @andrewsu and @newgene can weigh in.

---

Notes:

• lab Slack threads on bugs and ID-prefix handling

[colleenXu]

Also a consideration: do pending APIs (and Multiomics / Text-mining ones) need to keep both a ontology-specific field (where the value is probably not prefixed) and a general "ID" field for nodes?

Like here (the id vs HP field):

  "object": {
    "HP": "0000360",
    "id": "HP:0000360",
    "name": "Tinnitus",
    "type": "biolink:PhenotypicFeature"
  }

[colleenXu]

CC @tokebe @andrewsu @newgene

BTE previously automatically added the prefix to RHEA IDs, and at some point, stopped (this is the confusing part!)

I only noticed today while checking on another issue (primary knowledge source). And I tested / made this commit to add the needed prefix for operations using RHEA IDs as input: NCATS-Tangerine/translator-api-registry@d78fd51

I don't know the scope of this issue without doing a review of all the x-bte operations...my own musings below:

• the comments I left on (almost all?) yamls should help me find these cases where we relied on BTE automatically adding the prefix....or I could go through all the operations using https://github.com/biothings/biomedical_id_resolver.js/blob/main/src/config.ts#L4 ID-namespaces as input (dunno if output is an issue too?)
• is there any issue with using addPrefix()? using replPrefix would be safer (avoid cases of adding prefix twice like "CHEBI:CHEBI:1234")
  • this is an easy find-replace. looks like only two multiomics yamls are currently using addPrefix: drug-response and ehr risk

[colleenXu]

Noticed issue with GO prefixes and fixed by ensuring prefix is added to sub-queries NCATS-Tangerine/translator-api-registry@3703119

[colleenXu]

Noticed issue with MP prefix and fixed by ensuring prefix is added to sub-queries NCATS-Tangerine/translator-api-registry@72748a1