STRC h06 Hill Sensitivity Sweep 2026-04-23
Bounds analysis for the ⚠ CIRCULAR FIT Hill coupling (HILL_K=200, HILL_N=2, MAX_BOOST=3) flagged by STRC h06 Parameter Provenance Audit 2026-04-23. Answers: does Strategy A’s therapeutic claim survive Hill parameter uncertainty, or is it a point-fit artefact?
Spoiler: per-OHC claim survives (ROBUST); cochlea-mean claim fails architecturally — independent of Hill fit.
Method
Script: mrna_lnp_pkpd_hill_sensitivity.py. Fixed reference regimen (middle of prior sweep for clarity):
| Fixed | Value |
|---|---|
| modification | m1psi (mRNA t½ = 12 h in OHC) |
| interval | 28 d |
| dose_intra | 1000 mol/OHC |
| horizon | 365 d |
| threshold | tx_trough ≥ 2.0 fold |
Grid (5×5×5 = 125 points):
| Axis | Values |
|---|---|
| HILL_K | 50, 100, 200, 300, 500 |
| HILL_N | 1, 1.5, 2, 3, 4 |
| MAX_BOOST | 1.5, 2, 3, 4, 5 |
Per-OHC ODE integrated per point; cochlea_mean computed at 3 LNP efficiencies (0.8%, 5%, 20%) post-hoc via cochlea_mean = eff·tx + (1-eff)·1.
Results
Per-OHC robustness (transfected OHC only)
- 60% of grid (75/125 points) reaches
tx_trough ≥ 2.0 - Verdict: ROBUST — claim “per-OHC therapeutic at moderate mRNA dose” survives Hill uncertainty over the tested parameter space.
- Minimum MAX_BOOST for therapeutic, by (HILL_K, HILL_N): fails at HILL_K≥300 with HILL_N=1 (low cooperativity + weak affinity), but succeeds almost everywhere else at MAX_BOOST ≥ 3.
Cochlea-mean ceiling (population-level — what matters for hearing)
| LNP targeting | eff | Max cochlea_mean over grid | Theoretical ceiling at MB=5 | Therapeutic reached anywhere? |
|---|---|---|---|---|
| untargeted | 0.008 | 1.032x | 1.032x | ❌ No |
| cochlear_tropic | 0.05 | 1.20x | 1.20x | ❌ No |
| OHC_targeted | 0.20 | 1.80x | 1.80x | ❌ No |
Zero points in the 125-grid achieve cochlea-mean therapeutic (≥2x), at any LNP efficiency up to 20%.
The architectural identity
For any targeting efficiency eff and any Hill ceiling max_boost:
cochlea_mean ≤ eff · max_boost + (1 - eff)
= 1 + eff · (max_boost - 1)
For cochlea_mean ≥ 2, need:
eff · (max_boost - 1) ≥ 1
Case-by-case at realistic max_boost:
| max_boost | Min eff for therapeutic |
|---|---|
| 3 (base) | 0.50 (50% OHC transfection) |
| 5 (optimistic) | 0.25 (25%) |
| 10 (implausible) | 0.11 (11%) |
Current best reported LNP cochlear efficiency: 0.8% (untargeted, Yeh/Gao-class). Cochlear-tropic LNPs: ~5% hypothetical. OHC-targeted ligand-LNPs: 20% aspirational. None of these cross the threshold at biologically plausible MAX_BOOST.
This is not a Hill-fit problem — it is an architectural ceiling imposed by the additive population formula with (1-eff) OHCs contributing fold=1 to the mean.
Implications
For #6 Strategy A monotherapy
- Mechanism tier: 3 → 2. Splicing-boost mRNA therapy cannot reach cochlea-mean therapeutic under any Hill parameters at realistic LNP efficiencies. The circular-fit flag understates the problem; the fit is not even necessary — the ceiling holds regardless.
- Tier: A → B. axes now min(2, 2, 2) = 2 = B. Aligns with axes. Was generous at A.
For #4 Strategy B (linear, no Hill)
Strategy B is NOT bounded by this identity because it has no ceiling on tx fold. At high enough mRNA dose, s_exog/s_WT can be driven arbitrarily high, which lifts cochlea_mean = s_endo_frac + eff·(s_exog/s_WT) past 2x at any eff > 0.
However: achieving s_exog/s_WT ≥ 40x at per-OHC level (required to cross cochlea-mean 2x at 5% LNP) demands very high mRNA dose per transfected OHC and is limited by translation saturation, LNP payload, and tolerability — not by Hill fit. Strategy B’s earlier “stack adds 0.12-0.92% cochlea-mean at realistic AAV” stands as a modest contribution; is NOT the monotherapy path.
For #4 + #6 Stack
The “stack” intuition (A + B combined) is weaker than it appears: at realistic LNP, A contributes ≤ 0.20x cochlea-mean increment (even at MAX_BOOST=5, OHC-targeted 20%). The stack is ≈ B alone with a small Hill bonus. This does not invalidate Strategy B’s own claim; it reframes the stack as “B + marginal A” rather than “synergistic A+B.”
What this DOES NOT say
- Per-OHC correction (for the transfected 20% of OHCs at OHC-targeted LNP) is plausible. If hearing recovery can tolerate a 20% OHC subset being corrected at 3-5x, Strategy A might still help locally. But the population-mean metric — the relevant one for broadband hearing — cannot cross 2x.
- Does NOT rule out Strategy A as an adjuvant to #3 Mini-STRC AAV (Mini-STRC already integrates genetically — adjuvant mRNA is a different architecture).
- Does NOT invalidate the circular-fit flag — Hill fit is still circular; the bounds just show the circular fit is not the binding constraint.
Ranking delta
- #6 mRNA Therapy: A → B. Mech 3 → 2. Deliv 2, Misha 2 unchanged. Architectural ceiling, not Hill-fit dependency.
- #4 Strategy B: unchanged B. Stack claim reframed (“stack ≈ B + marginal A at realistic LNP”), not invalidated.
What could move it back up
- LNP OHC-targeting above ~30-40% efficiency (not currently published; would need novel ligand-LNP chemistry or intracellular-delivery breakthrough).
- Alternate amplification mechanism bypassing the additive
(1-eff)penalty — e.g., paracrine factor that boosts untransfected OHCs. Speculative; no mechanism proposed. - Primary RBM24-titration lit in OHCs: would replace circular fit with measured Hill params. Does NOT resolve architectural ceiling. (Still worth retrieving for defensibility.)
Connections
[part-of]STRC mRNA Therapy Hypothesis[part-of]STRC Hypothesis Ranking[see-also]STRC h06 Parameter Provenance Audit 2026-04-23[see-also]STRC mRNA-LNP Strategy B Full-Length[see-also]STRC AAV-LNP Stack PKPD[see-also]feedback_literature_first