GROUNDING — h05 calcium-oscillation literature audit

Weights: STATES · VARIANT (bridging derivation inline) · ABSENCE.


G1 — CaMKII autophosphorylation rate k_auto: A MEASUREMENT EXISTS

The 2026-04-25 audit recorded this parameter as ”⚠ not measured. Fitted; Chao 2011 is structural only.” That is wrong — the measurement exists and this run found it.

Exact query run (WebSearch):

CaMKII autophosphorylation rate constant k_auto Thr286 measured per second kinetics Bradshaw Meyer

The search engine iterated four sub-queries before landing on the primary. Final verbatim from the synthesis:

“The autophosphorylation rate at 30 degrees C was approximately 12 s(-1), which means the rate constant k_auto for CaMKII autophosphorylation at Thr286 is approximately 12 per second at 30°C.”

“This was measured using chemical quenched flow kinetics, and this rate was more than 10-fold faster than past estimates from earlier studies. The study also established that the autophosphorylation rate was independent of the CaMKII concentration, results corroborating intraholoenzyme activation.”

“The full reference for this work is: Bradshaw, Hudmon, and Schulman, ‘Chemical quenched flow kinetic studies indicate an intraholoenzyme autophosphorylation mechanism for Ca2+/calmodulin-dependent protein kinase II’ J. Biol. Chem., 277 (2002), pp. 20991-20998.”

PMID 11925447.

Verification attempt (WebFetch on https://pubmed.ncbi.nlm.nih.gov/11925447/): returned a cookie-consent interstitial, no abstract. The record was not independently confirmed against the primary text this run.

Evidential weight: VARIANT — a strong, precisely-specified lead, NOT a sourced value. The reference is internally consistent (real journal, real volume/page span, plausible PMID, mechanism matches the paper’s stated title) and the number is quotable. But it comes from a search engine’s restatement, not from a locally parsed PDF, and STRC AGENTS.md §0c rule 4 is explicit: “Never fabricate ‘estimates’ from training data. If it’s not in a locally-parsed paper, it doesn’t exist for modeling.” A search-snippet number is closer to that prohibition than to compliance.

So this is deliberately not promoted to ✅ and no sources/lit/ note is created from it. What changes is the blocker’s character: h05’s CaMKII row moves from “no kinetic measurement exists” (false) to “the measurement exists at PMID 11925447, ~12 s⁻¹ at 30 °C; fetch the PDF, parse with mineru -p <in> -o <out> -b vlm-engine, file the note, then close.” That is a one-step remedy replacing a dead end.

Caveat to carry into any use: 12 s⁻¹ is measured on brain αCaMKII holoenzyme at 30 °C. h05 models an OHC at body temperature. Isoform and temperature transposition must be stated when the value is adopted.


G2 — the three STRC-specific kinetic parameters: ABSENCE, cross-corroborated

Not re-searched this run. The 2026-04-25 audit ran an exhaustive pass; its findings are quoted here so this file stands alone:

  • STRC mRNA t½“No primary paper measuring STRC mRNA t½ in any cell type exists in PubMed or bioRxiv as of 2026-04-25.” The phantom “Sharma 2018” was retracted. Model values STRC_MRNA_DECAY_S = 1e-4 (t½ ~2 h) and k_mRNA_deg = 0.0004 (t½ ~30 min) are unsourced calibration parameters that disagree with each other ~4×.
  • STRC protein t½“No primary paper measures STRC protein half-life in OHCs or any mammalian system.” Cross-script values 38 h vs 30 days — a 20× internal inconsistency, itself evidence that neither is anchored.
  • STRC copy number per OHC“No published proteomics paper (including mass spectrometry surveys of stereocilia) quantifies STRC copy number per OHC.” The phantom “Krey 2015” resolves to Wilmarth et al. 2015 Sci Data (PMID 25977827), in which STRC is not quantified. target_protein = 15,000 is a calibration set-point.

Independent cross-hypothesis corroboration (new this run): h09’s audit chain reached the same conclusion from a different direction and a different dataset — Krey 2019 eLife (PMC6855842), all supplementary materials checked (MaxQuant output, expression heatmaps, 186-gene scRNA-seq list, antibody table), STRC absent from every one. See GROUNDING/hypc-h09-hydrogel.md §G5 and sources/lit/2026-04-23-krey-2019-stereocilia-proteomics-elife.md.

Two audits, two hypotheses, two Barr-Gillespie-lab datasets (2015 mouse vestibular, 2019 chick utricle), same answer. STRC abundance per hair cell has not been measured.

Evidential weight: ABSENCE, high confidence. These three are not “un-retrieved”; they are unmeasured. Only the wet-lab gates the prior audit named can produce them: actinomycin-D chase for mRNA t½, cycloheximide chase for protein t½, sorted-OHC single-cell proteomics for copy number.


G3 — PDE4 Vmax and CRE K_half: partial and unconfirmed

Carried from the 2026-04-25 audit, not re-searched:

  • PDE4 Km = 4.8 µM is genuinely sourced — Owens et al. 1997 Biochem J (PMID 9337850), full-length PDE4A, filed as sources/lit/1997-owens-pde4a-kinetics.md. The model’s PDE4_KM_NM = 4000 (4 µM) is consistent. ✅
  • PDE4 Vmax, OHC-specific, is a fitted model parameter with no primary source. Generic PDE4 Vmax values do exist in enzymology; an OHC-specific one does not. The honest framing is therefore split: retrievable as a generic enzyme constant, not-modelable as an OHC-specific one.
  • CRE half-saturation K_half — the cited “Cha 2010” PMID was never confirmed and the prior audit found no replacement. Status: unverified citation, open.

Evidential weight: MIXED — recorded as carried, not as re-established.


G4 — h05 values confirmed sourced by the prior audit (spot-checked, not re-derived)

parametervaluesource note in sources/lit/PMID
apical compartment volume V_apex~50 fL1998-lumpkin-hudspeth-stereocilia-calcium-regulation9698322
resting [Ca²⁺]20–310 nMsame9698322
Ca²⁺ buffer capacity100–1000×2005-hackney-calcium-buffering-proteins16120789
PMCA extrusion ratek ~30 /s1998-yamoah-pmca-stereocilia-extrusion + 2010-beurg-calcium-balance-ohc9425003, 20427623
PDE4 Km4.8 µM1997-owens-pde4a-kinetics9337850
CaM EC50, CaMKII holoenzyme6.0 µM (Hill n = 2.1)2011-chao-camkii-holoenzyme-structure21884935

Note on the last row: the model uses Kd_CaMKII = 1 µM, ~6× below the measured holoenzyme EC50. The prior audit flagged this in-script as a model choice. It is not a phantom, but it is a 6× departure from the one measured anchor and should not be described as “sourced.”