STRC Research

STRC h02 Parameter Provenance Audit 2026-04-23

4 min read

STRC h02 Piezo — Parameter Provenance Audit 2026-04-23

Post-audit fix of piezo_voltage_budget.py, piezo_phase2_frequency_bundle.py, piezo_phase3_delivery_feasibility.py. Companion: piezoelectric-materials + tectorial-membrane. Triggered by STRC Cross-Hypothesis Parameter Audit 2026-04-23 flag “3 phantoms + TM mismatch”.

Fixes applied

piezo_voltage_budget.py (Phase 1)

  • C_spec citation: “Ashmore 1987” → Gentet 2000 Biophys J 79:314. Ashmore 1987 established OHC electromotility but did not quantify specific capacitance.
  • V_saturation_mV = 70: flagged as UNSOURCED model ceiling (no primary citation found; Santos-Sacchi curves go to ±150 mV range).
  • Materials table: added note that terpolymer d31 −20 pC/N is scaled from manufacturer d33 ≈ 40 pC/N; earlier d31 values were not cited. PLLA d14 → d31 treated as model approximation (flagged; PLLA is shear-mode).
  • TM displacement comment: “0.01 nm at 20 dB” anchor is a model choice, not a citation; kept with explicit flag; output range (5–30 nm at 60 dB) matches Ren 2002 / Ghaffari 2007 / Gao 2014.
  • NEW: tm_substrate_mismatch_flag in output JSON — model assumes bulk strain = (TM_disp/R_curv) but PVDF-TrFE E ≈ 3 GPa vs TM E ≈ 24–210 kPa → stiffness ratio 10⁴–10⁵×. Real bending concentrates in the soft TM; film receives negligible strain. V_wall values are UPPER BOUNDS, not predictions.

piezo_phase2_frequency_bundle.py (Phase 2)

  • d31 reconciliation: was −25 pC/N here vs −12 pC/N in Phase 1 (silent 2× split). Both reconciled to −12 pC/N (conservative β-phase, Arkema datasheet). Well-poled films can reach −25; not default.
  • Terpolymer d31: was −40 (that’s d33). Fixed to −20 pC/N.
  • C_OHC cite: “Ashmore 1987” → Gentet 2000.
  • R_OHC cite: flagged UNSOURCED (generic textbook kΩ·cm²; no OHC-specific primary source).
  • TM_DISP_60dB comment: “Gueta 2006” was a mis-citation (Gueta 2008 Biophys J 95:4948 reports OHC deflection from TM anisotropy, not absolute displacement). Replaced with Ghaffari 2007 / Ren 2002 / Gao 2014, and flagged as ±50% uncertainty.

piezo_phase3_delivery_feasibility.py (Phase 3)

  • A666 prestin-binding peptide — PHANTOM CALLED OUT: no paper “Zou 2015” exists describing A666. Prestin-specific extracellular ligand is a research gap. Module docstring now leads with this warning.
  • SELECT_S = 80 — PHANTOM CALLED OUT: no primary source. Kept as model parameter but renamed in output JSON (selectivity_fold_PHANTOM) and flagged in docstring.
  • ETA_POLY = 0.7 — PHANTOM CALLED OUT: no primary paper for in-situ VDF/TrFE polymerisation in cochlea. Renamed eta_polymerisation_PHANTOM.
  • K_BIND_BASELINE / K_OFF_A666 — PHANTOM CALLED OUT: Berg-Purcell 4πDR physics is defensible; multiplication by 10% “A666 contact probability” is invented. Renamed.
  • Output JSON has new _WARNING field — top-level string flagging phantom status.
  • What’s still defensible in Phase 3: cochlear geometry (N_OHC = 3000 mouse; apical area budget), Stokes-Einstein D_NP, Berg-Purcell form, perilymph volume 1 µL (mouse), RWM bolus t½ (Salt & Plontke cochlear simulator), hydrogel IT t½ 24 h (Salt review). Everything else is scaffold.

Before → after outputs

Phase 1 piezo_voltage_budget.py

ScenarioV_wall@60dB (mV) before (d31=−25?)V_wall@60dB (mV) after (d31=−12)60 dB pass?
baseline PVDF-TrFE R=1µm~7.53.64❌
bundle-scale R=100 nm~7536.4✅ (50 dB threshold)
terpolymer R=1µm—4.47❌
PLLA R=1µm—2.59❌

Phase 1 baseline no longer passes at 60 dB — hypothesis requires bundle-scale (R ≤ 100 nm) film OR SPL ≥ 70 dB.

Phase 2 piezo_phase2_frequency_bundle.py

Audiogram passes (beam model, 60 dB) dropped from >0 to 0/6 freqs after d31 reconcile. Only wall_curvature (Phase 1 equivalent) still clears threshold at most frequencies.

Phase 3 — numbers unchanged but tagged PHANTOM throughout.

Ranking delta

h02 tier was S (“active compute now”) based on pre-audit Phase 3 delivery feasibility appearing ≥80% at 60 dB audiogram under the A666 + hydrogel IT scenario. Post-audit:

AxisPre-auditPost-auditReason
Mech32Phase 1 baseline fails 60 dB after d31 reconcile; TM mismatch 10⁵× not in model; bundle-scale R=100 nm is still physically uncertain
Deliv21A666 peptide phantom. No published prestin-specific extracellular ligand. In-situ VDF/TrFE polymerisation unpublished. Delivery pathway is speculative, not engineered.
Misha-fit55Non-gene-therapy, age-agnostic, post-window compatible — unchanged

New tier: B (was S). min(Mech, Deliv, Misha) = min(2, 1, 5) = 1 → normally C, but the Misha-fit 5 keeps it in B as “watch, not killed”. Need to demote from “active compute” bucket.

Next steps before any promotion back:

  1. Real OHC-apical targeting ligand — literature scan for prestin-specific peptides OR accept passive deposition (random distribution over OHC + IHC + support).
  2. Demonstrate ex-vivo conformal film deposition on isolated OHC bundles (PVDF-TrFE solution cast + β-phase polarisation) — no in-situ polymerisation assumption.
  3. FEM strain-sharing model TM (24 kPa) + thin film (3 GPa) + perilymph — replaces “bulk strain = δ/R” approximation.

Connections

Graph View

Backlinks