QC review (deterministic + agentic): judgment-layer findings, 2026-06-02

[turbomam]

Findings from a QC pass run during release prep: the default-profile ROBOT report, metpo-proposal-lint, and an agentic review of all 291 classes + 113 properties (judgment-layer issues the deterministic checks miss). Items below were verified against the Sheet unless marked (pending).

Verified concrete defects

• alkaphilic mislabeled (3 classes). METPO:1003002 / 1003004 / 1003005 have primary labels alkaphilic / obligately alkaphilic / facultatively alkaphilic (missing the "li"); their exact synonyms correctly say alkaliphilic. Ontology-native labels, so this is a genuine spelling fix (in the Sheet).
• NaCl delta high (1000482) range column transposed. range_max=8 but equivalent_class_formula is [>= 8] (a lower bound). Every other >X high bin puts the value in range_min. Residue near NaCl delta high equivalent class formula is inverted #357 (which fixed the formula direction, not the range column).
• Fermentation (1002005) mis-parented under respiration with a self-contradictory definition ("A respiration that … without using an external electron acceptor"). Fermentation is canonically distinct from respiration (cf. GO:0006113 vs GO:0009060). Needs reparenting + definition fix.
• Halophily thresholds incoherent. moderately halophilic (1000623) "≥0.2 M" sits below slightly halophilic (1000625) "0.3–0.8 M"; extremely halophilic uses % while others use M. Standard Kushner scale: slight 0.2–0.5 M, moderate 0.5–2.5 M, extreme 2.5–5.2 M. Curator-written numbers, likely wrong.

Modeling issues to triage (pending curator decision)

• OMP:0000087 ("facultative") mapped onto 4 distinct oxygen classes (1000605 facultatively anaerobic, 1000608 facultatively aerobic, 1000612 facultative oxygen preference) plus microaerotolerant (1000610) — a single source class cannot be the correct target for all; tolerance != facultative.
• extreme hyperthermophilic (1000721) is a sibling of hyperthermophilic (1000617) rather than its child.
• strictly anaerobic (1000611) may duplicate obligately anaerobic (its differentia restates the parent).
• Title-case label inconsistency in the metabolism subtree (Aerobic respiration, Fermentation, …) vs lowercase elsewhere.

Already tracked (corroborated, not new)

• capnophilic (1005021) wrong parent (oxygen preference; it's a CO2 requirement) — see Model biochemical tests and their outcomes: add an assay/test parent; relate outcomes to tests, not subclass #461/Audit and fix problematic IAO:0000115 definitions across classes and properties templates #377.
• indole/MR/VP test classes defined as "An assay…" under phenotype (process-vs-quality category drift) — Model biochemical tests and their outcomes: add an assay/test parent; relate outcomes to tests, not subclass #461.
• Exact-synonym overload (mesophilic etc. on multiple bins) — 'mesophilic' as exact synonym on 8 temperature-bin classes overloads the term #444.

Method note

METPO's custom ROBOT report profile masks ~4000 default-profile findings; #467 fixed the largest underlying cause (untyped owl:deprecated). Source-bound related synonyms are reified (owl:Axiom + IAO:0000119 source) and their duplication across bins is intentional (cross-system value mapping) — not a defect.