Add METPO predicate pairs for high-volume unresolved MetaTraits categories

[turbomam]

Problem

The in-sheet resolver (resolve-metatraits-in-sheets) can't map 14 composed trait categories because metpo-properties.tsv lacks predicate pairs for them. These 14 categories account for ~2.85M assertions out of 5.34M total (53%) in the MetaTraits NCBI species-level data — making this the single most impactful gap in KGX expressibility.

The resolver currently covers 77% of trait types but only 47% of assertions by volume, because the unresolved categories are the highest-frequency ones.

Assertion volume by unresolved category

Source: metatraits.ncbi_species_summary collection (54,654 species, 5,343,971 trait summary items).

Category	Unresolved cards	Species-level assertions
growth	27	1,104,805
enzyme activity	132	430,726
degradation	13	272,746
hydrolysis	13	271,774
reduction	4	196,983
produces	232	162,157
oxidation	6	153,029
aerobic growth	2	77,314
utilizes	80	75,998
carbon source	20	60,959
builds acid from	16	24,658
assimilation	29	13,085
energy source	6	1,035
nitrogen source	6	345
Total unresolved	586	~2,845,614 (53%)

For comparison, resolved composed categories:

Category	Assertions
electron acceptor	386,615
denitrification	115,944
respiration	78,549
fermentation	45,667
builds gas from	306
builds base from	14
aerobic catabolization	10
anaerobic catabolization	4
Total resolved composed	~627,109 (12%)

The remaining ~35% are base/boolean traits (gram stain, sporulation, oxygen preference, motility, etc.) which are resolved via biolink:has_phenotype.

What needs to happen

For each unresolved category, add a positive/negative predicate pair to src/templates/metpo-properties.tsv, following the existing pattern (e.g., ferments / does not ferment). Some categories like growth and enzyme activity may need modeling discussion before predicate assignment.

Priority order (by assertion volume)

1. growth — 1.1M assertions, needs "uses for growth" / "does not use for growth" (NOTE: predicates METPO:2000012/METPO:2000038 already exist but the resolver reports these cards as unresolved — likely a CHEBI object gap, not a predicate gap)
2. enzyme activity — 431K assertions, predicates exist (METPO:2000302/METPO:2000303) but resolution fails for 132 cards (missing CHEBI/GO objects in source data)
3. degradation — 273K, predicates exist (METPO:2000007/METPO:2000033) — similar CHEBI gap
4. hydrolysis — 272K, predicates exist (METPO:2000013/METPO:2000039) — similar CHEBI gap
5. reduction — 197K, predicates exist (METPO:2000017/METPO:2000044) — similar CHEBI gap
6. produces — 162K, predicates exist (METPO:2000202/METPO:2000222) — similar CHEBI gap
7. oxidation — 153K, predicates exist (METPO:2000016/METPO:2000042) — similar CHEBI gap
8. Lower-volume categories (aerobic growth, utilizes, carbon source, etc.)

Refined diagnosis: Many of these categories do have predicate pairs already. The resolution failure is primarily due to missing CHEBI object mappings for the composed substrates. The fix is likely a combination of:

• Adding CHEBI mappings for substrates that MetaTraits doesn't provide ontology CURIEs for
• Extending the resolver's object normalization to handle more source formats

Resolution table reference

• Current resolution table: data/mappings/metatraits_in_sheet_resolution.tsv
• Coverage report: data/mappings/metatraits_in_sheet_resolution_report.md
• Property template: src/templates/metpo-properties.tsv

Cross-references

• METPO predicated for Madin and bactotraits Knowledge-Graph-Hub/kg-microbe#458 — METPO predicates for Madin/bactotraits
• Use synonym mappings for data sources from 'source mappings' tab in METPO sheet Knowledge-Graph-Hub/kg-microbe#480 — synonym mappings from METPO sheet
• define quality metrics esp regarding how mappings are used in KG-Microbe #204 — quality metrics for mapping coverage