Review Ingredients

Current Dataset: 1,102 total ingredients

• 1,034 mapped ingredients (93.8%) with ontology IDs
• 68 unmapped ingredients (6.2%) - intentionally unmappable

When to Use This Skill

Decision Table:

IF newly mapped batch → Use interactive review
IF full dataset check → Use batch review
IF enrichment needed → Use auto-correct (P3/P4)
IF critical errors → Use batch review with P1 filter
IF synonym issues → Use validate_synonyms.py

Review Workflows

1. Interactive Review (Single Ingredient)

Use case: Verify a specific ingredient after manual curation

# Review by preferred term
PYTHONPATH=src python scripts/review_ingredient.py "sodium chloride"

# Review by ontology ID
PYTHONPATH=src python scripts/review_ingredient.py --id CHEBI:26710

# Review with auto-correction suggestions
PYTHONPATH=src python scripts/review_ingredient.py "glucose" --suggest-fixes

Output:

• Rich UI panel showing current mapping
• Validation results (P1-P4 issues)
• Suggested corrections with apply/skip options
• Ontology metadata comparison

2. Batch Review (All Ingredients)

Use case: Validate entire dataset, generate comprehensive report

# Review all mapped ingredients
PYTHONPATH=src python scripts/batch_review.py \
  --output reports/validation_$(date +%Y%m%d) \
  --format md,json,html \
  --threads 4

# Filter by priority
PYTHONPATH=src python scripts/batch_review.py --priority P1,P2

# Filter by ontology source
PYTHONPATH=src python scripts/batch_review.py --source CHEBI

# Limit for testing
PYTHONPATH=src python scripts/batch_review.py --limit 10 --dry-run

Output:

• validation_report.md: Human-readable summary with statistics
• validation_data.json: Machine-readable issues + corrections
• dashboard.html: Interactive sortable/filterable dashboard

3. Automated Correction (P3/P4 Safe Issues)

Use case: Auto-fix issues that don't require human review

# Dry-run to preview changes
PYTHONPATH=src python scripts/auto_correct.py --dry-run

# Apply safe corrections (P3/P4 only)
PYTHONPATH=src python scripts/auto_correct.py --apply

# Enrich chemical properties only
PYTHONPATH=src python scripts/auto_correct.py --apply --types chemical_properties

# Add missing synonyms from ontologies
PYTHONPATH=src python scripts/auto_correct.py --apply --types synonyms

Safe corrections (auto-applied):

• Enrich missing chemical properties from OLS
• Add missing synonyms from ontology
• Normalize CURIE formats
• Fix whitespace/case in labels

Unsafe corrections (manual review required):

• Change ontology ID
• Modify preferred term
• Merge duplicate records
• Change mapping quality level

4. Claude Code-Assisted Review

Use case: Interactive validation with Claude's assistance

# Use this skill
/review-ingredients

# Or invoke via Skill tool with specific ingredient
/review-ingredients "calcium chloride dihydrate"

Claude will:

1. Load ingredient record from YAML
2. Run validation via IngredientReviewer
3. Explain issues in plain language
4. Propose corrections with rationale
5. Apply fixes if user approves
6. Update curation_history

Validation Rule Catalog

Priority Levels

Rule Definitions

P1 - Critical Errors

Rule P1.1: Ontology Term Existence