AUDIT — h05-calcium-oscillation (Calcium Oscillation Acoustic Therapy)

  • Tier: C · mech 3 / deliv 4 / misha_fit 2
  • lit_audit before: partial (2026-04-25), 6 blockers listed
  • lit_audit after: partial, 6 blockers — reclassified, one materially advanced
  • Grounding: GROUNDING/hypc-h05-calcium-oscillation.md

Prior work read

  • hypotheses/h05-calcium-oscillation/index.md, log.md
  • phases/STRC h05 Parameter Provenance Audit 2026-04-23.md
  • phases/STRC h05 Parameter Provenance Audit 2026-04-25.md
  • parameters/calcium-oscillation.md

The 2026-04-25 audit is the strongest of the three partial audits: it killed four phantom citations, corrected four more wrong-paper attributions, and created seven paper notes. Its blocker list was already a proper YAML list. This pass adds classification and one substantive correction.

Parameter-provenance table

#ParameterValue in modelClaimed sourceLocal sources/lit/ fileStatus
1Apical compartment volume V_apex~50 fLLumpkin & Hudspeth 1998, PMID 96983221998-lumpkin-hudspeth-stereocilia-calcium-regulation.md
2Resting [Ca²⁺]20–310 nMsamesame
3Ca²⁺ buffer capacity100–1000×Hackney 2005, PMID 161207892005-hackney-calcium-buffering-proteins.md
4PMCA extrusion rate~30 /sYamoah 1998 PMID 9425003 + Beurg 2010 PMID 20427623both filed✅ (as model estimate)
5PDE4 Km4 µM (PDE4_KM_NM = 4000)Owens 1997, PMID 9337850 (4.8 µM)1997-owens-pde4a-kinetics.md
6CaM EC50, CaMKII holoenzyme6.0 µM, Hill n = 2.1Chao 2011, PMID 218849352011-chao-camkii-holoenzyme-structure.md✅ (measured)
7Kd_CaMKII1 µMmodel choice🔶 6× below the measured #6; flagged in-script
8n_channels~134Fettiplace & Kim 2014, PMID 24987009✅ (derived: 2 per tip link × 60–100 stereocilia)
9STRC mRNA t½1e-4 /s (~2 h) and k_mRNA_deg = 0.0004 (~30 min)“Sharma 2018” — PHANTOM, retracted❌ NOT MEASURED; 4× cross-script mismatch
10STRC protein t½38 h (pivot) vs 30 days (rbm24)none❌ NOT MEASURED; 20× cross-script mismatch
11STRC copy number / OHC15 000”Krey 2015” — PHANTOM, retracted❌ NOT MEASURED
12CaMKII k_autofitted”Chao 2011 structural only”🔶 MEASUREMENT LOCATED THIS RUN — see below
13PDE4 Vmax (OHC)fittednone❌ no OHC-specific source
14CRE half-saturation K_half“Cha 2010” — PMID never confirmed❌ unverified citation, no replacement
15AC1 Vmax (cell-normalized)30–60× membrane-prep valuevolume-normalisation argument🔶 model choice, justified in-script
16CREB dephosphorylation k2–4× faster than Gonzalez & Montminy 1989no OHC source🔶 mismatch documented

Retrieval performed this run

targettoolexact queryoutcome
CaMKII k_autoWebSearchCaMKII autophosphorylation rate constant k_auto Thr286 measured per second kinetics Bradshaw MeyerHIT — Bradshaw, Hudmon & Schulman 2002 JBC 277:20991–20998, PMID 11925447, ~12 s⁻¹ at 30 °C by chemical quenched-flow
same, verificationWebFetchpubmed.ncbi.nlm.nih.gov/11925447/cookie-consent interstitial; abstract not retrieved — not independently confirmed

The one substantive correction

The 2026-04-25 audit recorded CaMKII k_auto as ”⚠ not measured. Fitted; Chao 2011 is structural only.” That is wrong. The measurement exists: Bradshaw, Hudmon & Schulman 2002 measured Thr286 autophosphorylation by chemical quenched-flow at ~12 s⁻¹ at 30 °C, more than 10× faster than the earlier estimates the field had been carrying, and showed the rate is independent of CaMKII concentration (the intraholoenzyme result).

The prior audit reached “not measured” because it searched around the structural CaMKII literature (Chao 2011) rather than the kinetics literature. This is the characteristic failure of a phantom-hunting pass: once a citation is proven false, it is easy to conclude the value behind it does not exist, when in fact only the citation was wrong.

I am not marking this ✅. The number came from a search engine’s restatement, the PubMed verification fetch failed, and §0c rule 4 is explicit that a value not in a locally-parsed paper does not exist for modeling. Promoting a search snippet to a sourced constant is precisely the failure this discipline exists to prevent.

What changes is the blocker’s character: from “no kinetic measurement exists” (a dead end that invites nothing) to “the measurement is at PMID 11925447, ~12 s⁻¹ at 30 °C — fetch, MinerU, file the note, close it” (a half-hour task). Carry the caveat when it is adopted: brain αCaMKII holoenzyme at 30 °C, being transposed onto an OHC at body temperature.

Cross-hypothesis corroboration on STRC abundance

h05 blocker #11 (STRC copy number per OHC) and h09 blocker #2 (STRC_NORMAL_OHC_M = 1 µM) are the same missing measurement reached from two independent directions, and both audits independently concluded it is absent:

h05 routeh09 route
dataset checkedWilmarth 2015 Sci Data, PMID 25977827 (mouse vestibular stereocilia proteome)Krey 2019 eLife, PMC6855842 (chick utricle single-cell proteomics + scRNA-seq)
what was checkedMS survey of stereocilia proteinsMaxQuant output, expression heatmaps, 186-gene list, antibody table
resultSTRC not quantifiedSTRC absent from all

Two hypotheses, two Barr-Gillespie-lab datasets, same answer. The convergence raises confidence that this is a real literature absence rather than two independent retrieval failures — and it means one wet-lab measurement (sorted-OHC single-cell proteomics) would close a blocker in both h05 and h09. That is worth knowing when the wet-lab budget is allocated; it is recorded in RANKING-DELTA.md.

Why partial and not not-modelable

h05’s blockers split cleanly:

  • Not-modelable class (3): STRC mRNA t½, STRC protein t½, STRC copy number. Genuinely unmeasured anywhere; only wet-lab produces them. Note also that the 4× and 20× cross-script inconsistencies are themselves evidence of absence — values with a real anchor do not drift 20× between two scripts in the same hypothesis.
  • Retrievable class (3): CaMKII k_auto (lead in hand), PDE4 Vmax (generic enzyme constant exists; OHC-specific does not), CRE K_half (citation unverified, replacement not yet sought).

Because half the blockers are retrievable — and one is a PMID away — the hub cannot honestly be not-modelable. partial is the correct state and, unlike h09’s, it is a partial that further searching will actually reduce.

Verdict

lit_audit: partial · lit_audit_date: 2026-08-21 · 6 blockers

lit_audit: partial
lit_audit_date: 2026-08-21
lit_audit_blockers:
  - "STRC mRNA t1/2 — NOT MEASURED in any cell type (PubMed + bioRxiv exhaustive, 2026-04-25). 'Sharma 2018' was a phantom, retracted. Model carries two mutually inconsistent unsourced values: STRC_MRNA_DECAY_S = 1e-4 (t1/2 ~2 h) and k_mRNA_deg = 0.0004 (t1/2 ~30 min), a 4x mismatch. NOT-MODELABLE class: only an actinomycin-D chase in an OHC line can produce it."
  - "STRC protein t1/2 — NOT MEASURED in OHCs or any mammalian system. Cross-script values 38 h (pivot) vs 30 days (rbm24) differ 20x, which is itself evidence that neither is anchored. NOT-MODELABLE class: cycloheximide chase + immunofluorescence in OHC required."
  - "STRC copy number per OHC — NOT MEASURED. 'Krey 2015' phantom resolves to Wilmarth 2015 Sci Data (PMID 25977827), in which STRC is not quantified; target_protein = 15000 is a calibration set-point. CROSS-CORROBORATED 2026-08-21 with h09: Krey 2019 eLife (PMC6855842) all SI independently checked, STRC absent from every dataset. Same missing measurement as h09's STRC_NORMAL_OHC_M blocker — one sorted-OHC single-cell proteomics run would close both. NOT-MODELABLE class."
  - "CaMKII k_auto — MEASUREMENT LOCATED 2026-08-21, note not yet filed. The 2026-04-25 audit's 'not measured, Chao 2011 is structural only' was WRONG: Bradshaw, Hudmon & Schulman 2002 J Biol Chem 277:20991-20998 (PMID 11925447) measured Thr286 autophosphorylation by chemical quenched-flow at ~12 s^-1 at 30 C, >10x faster than prior estimates, rate independent of CaMKII concentration. Value came from a search-engine restatement and the PubMed verification fetch returned a cookie interstitial, so it is NOT promoted to sourced per AGENTS.md 0c rule 4. Remedy: fetch the PDF, parse with 'mineru -p <in> -o <out> -b vlm-engine', file sources/lit/2002-bradshaw-camkii-autophosphorylation-kinetics.md, then close. Caveat on adoption: brain alphaCaMKII holoenzyme at 30 C, transposed onto an OHC at body temperature."
  - "PDE4 Vmax (OHC-specific) — no OHC-specific primary source; currently a fitted model parameter. Split verdict: generic PDE4 Vmax IS measurable enzymology and retrievable; an OHC-specific Vmax is not measured anywhere. PDE4 Km is separately fine at 4.8 uM (Owens 1997, PMID 9337850, filed at sources/lit/1997-owens-pde4a-kinetics.md)."
  - "CRE half-saturation K_half — citation 'Cha 2010' unverified; the PMID was never confirmed and the 2026-04-25 pass found no replacement. RETRIEVABLE class, but not attempted this run (budget went to CaMKII, which was the higher-value correction). Next step: identify a primary CRE-reporter dose-response paper and file it, or record a confirmed absence."

Ranking delta

C-tier HELD. See RANKING-DELTA.md.