Editorial: Add therapeutic_agent CHEBI qualifiers to pharmacotherapy treatments

[claude]

Origin

From review comments on PRs #152 and #153 by @cmungall:

PR #153 (Osteoporosis):

"the generic pharmacotherapy terms need chebi qualifiers/postcomps"
"needs qualifiers" (Denosumab)
"needs chemical qualifiers" (Teriparatide)

PR #152 (Atrial Fibrillation):

"use post-composition for pharmocotherapy terms"
"indicate role" (Anticoagulation)
"use post-composition" (Rate Control)

Problem

Multiple treatments in these PRs use MAXO:0000058 (pharmacotherapy) as the treatment term without specifying the therapeutic_agent CHEBI qualifiers. Per the schema design, when using generic MAXO terms like pharmacotherapy, the specific drug should be captured via therapeutic_agent with CHEBI terms.

Affected Treatments

PR #153 - Osteoporosis

• Denosumab (RANKL inhibitor) - needs CHEBI term for denosumab
• Teriparatide (PTH analog) - needs CHEBI term for teriparatide
• Romosozumab (Sclerostin inhibitor) - needs CHEBI term for romosozumab

PR #152 - Atrial Fibrillation

• Anticoagulation (warfarin, DOACs) - needs CHEBI terms for warfarin and/or DOACs
• Rate Control (beta blockers, CCBs, digoxin) - needs CHEBI terms for representative drugs
• Rhythm Control (amiodarone, flecainide, sotalol) - needs CHEBI terms

Expected Pattern

From Wilson's Disease example in the codebase:

  treatment_term:
    preferred_term: pharmacotherapy
    term:
      id: MAXO:0000058
      label: pharmacotherapy
    therapeutic_agent:
      - preferred_term: zinc acetate
        term:
          id: CHEBI:62984
          label: zinc acetate

Proposed Action

1. Look up appropriate CHEBI terms for each drug
2. Add therapeutic_agent arrays to treatment_term blocks in both PR files
3. Validate with just validate and just validate-terms

[github-actions]

Issue Analysis Summary

This is an editorial enhancement task stemming from PR review feedback on #152 (Atrial Fibrillation) and #153 (Osteoporosis). The issue tracks the need to add CHEBI (Chemical Entities of Biological Interest) term qualifiers to pharmacotherapy treatments that currently only use the generic MAXO:0000058 (pharmacotherapy) term.

Context & Related Issues

This issue is directly related to #146 ("Define pattern for using the therapeutic_agent slot on TreatmentDescriptor"), which established the need for this capability. The pattern has now been implemented in the schema and demonstrated in the Wilson's Disease entry (lines 228-232 in kb/disorders/Wilsons_Disease.yaml).

Schema Design Pattern

The schema supports post-composition via the therapeutic_agent slot on TreatmentDescriptor:

treatment_term:
  preferred_term: pharmacotherapy
  term:
    id: MAXO:0000058
    label: pharmacotherapy
  therapeutic_agent:
    - preferred_term: zinc acetate
      term:
        id: CHEBI:62984
        label: zinc acetate

Key schema facts (from src/dismech/schema/dismech.yaml):

• therapeutic_agent is a multivalued slot (line 343), allowing multiple drugs per treatment
• Range is ChemicalEntityDescriptor (line 341), which binds to CHEBI terms
• Comments suggest: "Prefer CHEBI terms for specific drugs" and "Can also use NCIT for drug classes when CHEBI term unavailable" (lines 1101, 345-346)

Scope of Work

The issue identifies 9 treatments across 2 PRs that need CHEBI qualifiers:

PR #153 - Osteoporosis (3 treatments)

1. Denosumab - RANKL inhibitor monoclonal antibody
2. Teriparatide - PTH analog peptide
3. Romosozumab - Sclerostin inhibitor monoclonal antibody

PR #152 - Atrial Fibrillation (6 treatment categories)

4. Anticoagulation - warfarin, DOACs (direct oral anticoagulants)
5. Rate Control - beta blockers, calcium channel blockers, digoxin
6. Rhythm Control - amiodarone, flecainide, sotalol

Implementation Considerations

For monoclonal antibodies (denosumab, romosozumab):

• CHEBI includes biological macromolecules, but coverage for therapeutic mAbs may be limited
• May need to check if CHEBI has these or fall back to NCIT (NCI Thesaurus) which has extensive drug coverage

For drug classes (beta blockers, CCBs, DOACs):

• Issue description suggests "representative drugs" approach
• Alternative: Use CHEBI drug class terms if available (e.g., CHEBI:35522 for beta-adrenergic antagonist)
• Could list multiple specific agents per treatment

Validation requirements:

• just validate - schema conformance
• just validate-terms - ontology term validation against authoritative sources (critical anti-hallucination check)

Current Usage in KB

CHEBI terms are already used in at least 4 disorder files:

• Wilson's Disease (zinc acetate)
• G6PD Deficiency (3 chemical triggers)
• Alzheimer's Disease (5 chemicals)
• Peroxisome Biogenesis Disorder (2 chemicals)

Recommended Next Steps

1. Look up CHEBI IDs for each drug/drug class using OAK:

   uv run runoak -i sqlite:obo:chebi search "warfarin"

2. For biologics, verify CHEBI coverage or consider NCIT alternatives

3. Decide on specificity strategy for drug classes (representative drugs vs. class terms)

4. Update the YAML files in both PRs with therapeutic_agent blocks

5. Validate using the validation stack to ensure term accuracy

Priority & Impact

This is a data quality enhancement that improves semantic precision by:

• Making pharmacotherapy treatments machine-readable with specific drug identifiers
• Enabling computational queries like "find all disorders treated with anticoagulants"
• Following best practices for ontology-based knowledge representation
• Aligning with the pattern established in issue Define pattern for using the therapeutic_agent slot on TreatmentDescriptor #146

The affected PRs are likely blocked or will need updates, making this moderately time-sensitive.