STRC Research

STRC AC1-CREB Parameter Robustness

3 min read

STRC AC1-CREB Parameter Robustness

One-at-a-time local sensitivity sweep of the AC1-CREB sound-response ODE: 8 pharmacology parameters × (0.5×, 2×) = 16 perturbations. Only 1 breaks the hypothesis (SPL_HALF_SATURATION=140 dB, non-physiological). Robustness index 0.91 — the “chronic sub-loud dose” mechanism is not a narrow sweet-spot artifact.

Tornado table: fold silence → 45 dB under each perturbation

perturbationfold45Δ vs baselineflag
baseline1.82×—✅
K_TXN_MAX_FOLD 2× (→12)2.64×+45%best case
K_CA_AC1_NM 2× (→300 nM)1.93×+6%
AC1_VMAX 0.5×1.83×0%
PDE4_VMAX 2×1.85×+2%
K_CAMP_PKA 2×1.85×+2%
K_CREB_DEPHOS 2×1.83×0%
CA_CEILING 0.5× or 2×1.82×~0%irrelevant
SPL_HALF 0.5× (→35 dB)1.82×0%
K_CA_AC1 0.5× (→75 nM)1.37×−25%asymmetric
PDE4_VMAX 0.5×1.57×−14%
K_CAMP_PKA 0.5×1.57×−14%
K_CREB_DEPHOS 0.5×1.73×−5%
K_TXN_MAX_FOLD 0.5× (→3)1.34×−26%weak
SPL_HALF 2× (→140 dB)1.05×−42%⚠ BREAKS

Five inferences

  1. Pharmacology is not the bottleneck. None of the 7 intrinsic biochemistry parameters (AC1, PDE4, PKA, CREB) breaks the hypothesis at 0.5× or 2×. Monotonicity and >1.2× fold survive across the plausible range. The cascade topology itself carries the signal; specific rate constants are not fragile.

  2. Load-bearing knob: K_TXN_MAX_FOLD (CRE maximal induction). 3× → 1.34× fold, 6× → 1.82×, 12× → 2.64×. Literature values for CRE responsiveness span 3–20× depending on promoter and cell type. Real STRC induction likely 1.3–2.6×. Wet-lab pinning this parameter is the single highest-value experiment.

  3. Fragile at silence baseline, not at stimulus peak: halving K_CA_AC1 (→75 nM) reduces fold-change because silence Ca (25 nM) becomes 1/3 of Km instead of 1/6, so silence AC1 is already ~10% instead of ~3% → silence cAMP/protein already elevated → less headroom. Asymmetric sensitivity — the calibration of the silence baseline matters more than the stimulated peak.

  4. Ca drive ceiling is irrelevant — CA_CEILING_NM changes 1500–6000 nM leave the output unchanged. Bottleneck is downstream (CREB-P plateau at ~0.7). Any OHC Ca model above ~500 nM saturates the cascade equally well. No need to pin OHC apical [Ca²⁺] precisely.

  5. Only failure mode: SPL_HALF_SATURATION = 140 dB (i.e., cochlea barely activates until 140 dB). Non-biological; published Dallos 1978 and modern OHC-MET Ca calibration put half-sat at 50–70 dB SPL. So this “break” is not a real risk.

Implications for Touch Grass design

  • The “chronic sub-loud acoustic dose” strategy survives parameter uncertainty. No fragile sweet spot.
  • Minimum target dose: 45 dB, 1-3 h/day, chronic
  • Saturating dose: 60 dB (no further benefit above this — cascade saturates)
  • Expected STRC protein fold: 1.3× – 2.6× (depending on actual CRE responsiveness)
  • Primary readout: pCREB Ser133 IHC (48× fold, easy to detect) — far better experimental target than STRC protein (2× fold, hard)

Wet-lab priority order

  1. Measure STRC CRE responsiveness: basal vs forskolin-induced STRC transcript fold in HEI-OC1 or hair-cell organoids. Pins K_TXN_MAX_FOLD, removes the last major parameter ambiguity.
  2. pCREB IHC ± sound exposure (see STRC AC1-CREB Monotonic Sound Response).
  3. STRC mRNA qPCR from cochlear explants ± sound.

Phase 3 questions (not yet computed)

  • Chronic adaptation: does 6-month daily dosing trigger PDE4 upregulation → loss of effect? Needs PDE4 feedback module in the ODE.
  • Diurnal variation: does CREB-P circadian gating alter effective daily dose?
  • Species-specific K_TXN: need mouse STRC-CRE reporter assay to pin the load-bearing parameter.

Files / Models

  • ~/STRC/models/ca_oscillation_phase2_sensitivity.py — OAT sweep script
  • ~/STRC/models/ca_oscillation_phase2_sensitivity.json — full perturbation table
  • ~/STRC/models/ca_oscillation_phase2_sensitivity.png — tornado chart + dose-response curves

Connections

Graph View

Backlinks