update BTE to biolink-model 2.4.8

[colleenXu]

Multiomics EHR risk kp api needs to have 2.4.8 since it has new predicates it needs to use

Process to update biolink (mostly @tokebe, some me)

• Dev makes PR in biolink-model repo
• CX Reviews 2.4.8 (differences from current 2.2.13)
• basic testing of PR
• merging it to main branch + deploying it

Parallel process (mostly me)

• Updating Multiomics EHR risk kp api x-bte annotation
• Updating x-bte annotation if needed for the 2.4.8 change...

[colleenXu]

https://github.com/biolink/biolink-model/releases/tag/v2.4.8

[tokebe]

PR here

[colleenXu]

Differences between 2.2.13 (current) and 2.4.8 (upgrading-to) that we might care about (mostly for CX to deal with)

Predicates:

• deprecated
  • biolink:has_real_world_evidence_of_association_with (line 1733)
  • biolink:approved_to_treat (line 4822; so its inverse is deprecated too)?
• added
  • biolink:associated_with_likelihood_of (starts line 1745) / biolink:likelihood_associated_with. Likely won't make any x-bte annotation to this; instead use the more specific children terms
  • biolink:associated_with_increased_likelihood_of (starts line 1758) / biolink:increased_likelihood_associated_with. For multiomics ehr risk kp api
  • biolink:associated_with_decreased_likelihood_of (starts line 1758) / biolink:decreased_likelihood_associated_with. For multiomics ehr risk kp api
  • biolink:target_for (starts line 1778) / biolink:has_target: specifically for a gene-disease relationship (A gene is a target of a disease when its products are druggable and when a drug interaction with the gene product could have a therapeutic effect)
    • could use with the edge-attribute "has_evidence" where the value is something in the druggable_gene_category_enum (aka DrugCentral / Pharos stuff; starts line 244) (starts line 10629)
  • biolink:assesses (starts line 2014) / biolink:is_assessed_by: like a chemical's affect was assessed in an assay on a protein/tissue/phenotype/organism/cell-line....
• inverse added:
  • biolink:has_active_component (starts line 1807): inverse of active_in, used to relate a CellularComponent to a GeneOrGeneProduct
  • biolink:has_predisposing_factor (starts line 4587): inverse of predisposes
  • biolink:is_ameliorated_by (starts line 4717): inverse of ameliorates
  • biolink:is_exacerbated_by (starts line 4742): inverse of exacerbates
  • biolink:occurs_in_disease (starts line 6350)
• changed spelling:
  • has_capability (was capability_of before...; line 5705)
• usage:
  • use part_of to relate MolecularActivity (aka reaction) to Pathway it participates in

SEMMEDDB mappings:

• predicates:
  • biolink:associated_with mapped to SEMMEDDB:associated_with (line 1522)
  • biolink:interacts_with mapped to SEMMEDDB:interacts_with (line 2049; was mapped before to physically_interacts_with)
  • biolink:diagnoses mapped to SEMMEDDB:diagnoses (line 2246; was mapped before to has_biomarker)
  • biolink:predisposes mapped to SEMMEDDB:predisposes (line 4585)
  • broad-mapping of biolink:exacerbates to SEMMEDDB:complicates (line 4740)
• node categories / semantic types
  • narrow-mapping of ChemicalEntity to STY:T129 (aka imft / Immunologic Factor) (line 7997)
  • narrow-mapping of SmallMolecule to a bunch of things (starts line 8024): bacs / hops / horm / inch / orch / carb / eico / lipd / nsba / opco / strd / vitamin

Prefixes:

• INXIGHT (https://drugs.ncats.io/)? ncats.drug: 'https://drugs.ncats.io/drug/' (line 84). Added as a prefix for ChemicalEntity (line 8005), MolecularMixture (line 8185), Drug (line 8396), Protein (line 9000)
• bioplanet (already changed biothings bioplanet yamls to use this prefix): ncats.bioplanet: 'https://tripod.nih.gov/bioplanet/detail.jsp?pid=' (line 103). Added as a prefix for Pathway (line 8317)
• CPT is a prefix for Procedure (line 7785)

Node or edge-attributes:

• deprecates:
  • biolink:source (line 738)
  • biolink:original_knowledge_source (line 6770)
• adds:
  • biolink:has_chemical_role (lines 1289-1294). But the spelling is confusing...line 7195 introduces a chemical_role as an attribute
  • biolink:routes_of_delivery (lines 1348-1353) but gives a specific set of values (drug_delivery_enum, in lines 278-283): inhalation, oral, absorbtion through the skin (TYPO), intravenous injection
  • multiomics ehr risk kp api attributes? (starts line 7046) "supporting study"
• biolink:provided_by as a node-attribute to say which knowledge-provider created/assembled node and its attributes (line 6754)

---

Other (probably not important to us right now, but good to know)

• Starting to add "opposite" relationships for predicates/attributes (node ones??) using biolink:opposite_of in the predicate's annotations section
• CHEMBL.MECHANISM:substrate related-mapping to biolink:increases_degradation_of (line 3135). Hmm...
• TYPO line 3169 value: increases degredation of
• deprecates node categories / semantic types GeneOntologyClass (line 7297), UnclassifiedOntologyClass (line 7300), Nutrient (line 8415; and its children Macronutrient and Micronutrient), Vitamin (line 8436)
• MISTAKE line 9756: original knowledge source added as a "slot" even though it was deprecated?

[colleenXu]

Questions for @tokebe:

• how do we handle deprecated stuff (property under the predicate or node category/semantic-type, not removed from yaml yet). I notice that often the hierarchy (is_a) is removed, so that might lead to some "undefined" predicates??
• does BTE strip any unintentional extra white-space from the x-bte annotations (think predicates and input/output IDs and types...)? sometimes I add an extra space on accident....and it's super tedious to go through all x-bte annotations to try to fix this if it's an issue... >.<
• do we have code regarding edge-attributes and biolink:original_knowledge_source? I think we don't (update edge provenance info to comply with Translator Standard -- July 1 #208 (comment), CTRL-F "original" since the posts in this issue are lengthy)....but I think this is a quick check...

[tokebe]

• It would appear that BTE doesn't take into account the deprecated property in the biolink-model package, nor seemingly anywhere relating to the biolink model. I'm not sure I follow on how this might lead to undefined predicates, could you be more specific?
• If you mean trailing whitespace specifically in the yamls...I'm reasonably sure that this kind of whitespace is trimmed when yaml is converted to JSON, which (if I understand the process) happens before BTE even retrieves the smartapi specs?
  • I understand that fixing every instance would require going back through every file, and isn't terribly necessary given it doesn't present an issue, however I do recommend an auto-formatter for general use to avoid this going forward.
• I don't see any hardcoded mentions of biolink:original_knoweldge_source across the workspace.

[colleenXu]

"Deprecated": Actually....oops this shouldn't be a problem. I think it's fine that we don't take into account deprecated. It's more for curators/x-bte annotation writers and editors....to know that we shouldn't use those predicates and node-category/semantic-types anymore...

• My worry was that if a predicate or node-category/semantic-type is deprecated AND its is_a property is removed, BTE will put the predicate/node-category as "biolink:undefined" and may have buggy behavior.
  • I think this has happened in the past, but with predicates/node-category stuff that wasn't in the biolink-model at all....so a slightly different situation
  • I suspect this may also happen when the is_a property is removed, since then BTE may not know how to place the term in the hierarchy of predicates/node-categories
• But now that I check:
  • has_real_world_evidence_of_association_with (line 1733) still has an is_a property
  • approved_to_treat (line 4822) still has an is_a property
  • Vitamin (line 8436) still has an is_a property

[colleenXu]

2nd point on trailing white-space on yamls:

• yep that was my worry. I don't see any trailing whitespaces ( " or " ) in BTE's data/predicates.json. I do see some in the smartapi_specs.json but not for x-bte annotation (for descriptions. although only a few are true issues for formatting. Others are legit).
• I'm not aware of how to use an auto-formatter for yaml....I only just found one for Markdown (github I think)...

[colleenXu]

And 3rd point, you misspelled biolink:original_knowledge_source....I presume you did search correctly :P

[tokebe]

I actually searched for knowledge_source more generally because I recalled some instances of that, but none of them had original_. But yes, I just typed my answer quickly :)

[colleenXu]

Done. Updated predicates for MyChem, MyGene, BioThings SEMMEDDB: see commits NCATS-Tangerine/translator-api-registry@d02f89a and NCATS-Tangerine/translator-api-registry@40000c3

and refreshed their registrations afterwards.