Ototopical self-assembling peptide hydrogel hypothesis for STRC-null hearing loss. The therapeutic idea is to form a synthetic horizontal top connector at the stereocilia surface using a RADA16-WH2-STRC tail91 peptide, restoring mechanical coupling without AAV, mRNA, cells, or genome editing.
Promoted after Phase 4m closed the conservative avidity gate.
Phase 4m F-actin-adjusted WH2 Kd soft floor used in the conservative avidity model.
Phase 4n ideal Monte Carlo pass probability; 59.0% under a 30x entropic haircut.
Phase 4e modeled lower-to-upper ototopical dosing window before the toxic floor.
We are not asking whether h09 is ready for humans. We are asking whether a synthetic peptide hydrogel that substitutes for missing STRC horizontal top connector mechanics is coherent enough to justify a minimum feasibility package.
The five-step package below is intended as the smallest defensible experiment that could either earn h09 a real wetlab program or kill it cleanly.
Two columns kept symmetric on purpose: every claim is paired with a non-claim so the reviewer can see exactly where the dossier stops.
H09 is not a gene therapy. It is a local material intervention: deliver a peptide, let it assemble near the stereocilia surface, and use multivalent weak contacts to replace enough STRC top-connector mechanics to matter.
The hypothesis is mutation-agnostic for STRC-null patients if bundle-scale coupling can be restored.
One peptide combines actin engagement, self-assembly, and TMEM145-facing STRC tail mimicry.
The TMEM145 side moved from a generic tail premise to a specific modeled contact surface.
Single-site F-actin binding is weak, but multivalent presentation can make it useful.
The biological target is partial mechanical coupling rescue, not full restoration of native STRC architecture.
Promotion to S-tier came only after the weak monovalent WH2/F-actin question was converted into a calibrated avidity model. Phase 4k is shown as a caution; it does not carry the main claim.
| Phase | Result | Evidence | Decision |
|---|---|---|---|
| Phase 1 | Self-assembly screen | 5/6 candidate scaffolds passed the initial assembly filter; RADA16-WH2-native became the top track. | Advance self-assembling peptide scaffold rather than soluble protein replacement. |
| Phase 3 | Tail91 retool | Correct-epitope STRC tail91 reached TMEM145 ipTM 0.68 and actin ipTM 0.51. | Lock the 134-aa PEPTIDE_TAIL91 design for Phase 4 stress testing. |
| Phase 4b v2 | Developability scrub | Cys11/52->Ser candidate held TMEM145 median ipTM 0.55 and actin median ipTM 0.51 over n=5 each. | Production candidate can remove oxidation liabilities with modest TMEM145 cost. |
| Phase 4j | Avidity model | Original conservative Kd_mono 10 mM, C_eff 50 mM, N=4 gave Kd_eff 80 uM, failing by 1.6x; nominal case passed at 31 nM. | The hypothesis needed a defensible Kd_mono floor, C_eff proof, or higher-valency geometry. |
| Phase 4n | Monte Carlo stress test | P(PASS ideal)=0.803; P(PASS 30x entropic haircut)=0.590; Kd_mono was the dominant uncertainty. | Prioritize WH2/F-actin affinity calibration over extra structural storytelling. |
| Phase 4m | Kd soft-floor calibration | WH2 calibration fit R2=0.690; h09 predicted Kd_F=1.48 mM after 10,000x G-to-F scaling. | Re-run conservative avidity gate: N=3, C_eff 50 mM gives Kd_eff 1.30 uM, passing the therapeutic threshold. |
| Phase 4k | Fibril geometry caution | Tandem proxies reached four geometric WH2-actin contacts, but zero confidence-supported contacts across five AF3 jobs. | Keep geometry as a caution, not as the S-tier promotion evidence. |
| Phase 4p | Orthogonal WH2 controls | Boltz-2 placed h09 WH2 at ipTM 0.706; all negative controls were lower, while known positives remained higher or nearby. | Supportive orthogonal check; not a replacement for wet WH2/F-actin binding data. |
| Construct | Composition | Computational evidence | Role |
|---|---|---|---|
| tail91 v0 | WH2-RADA16-STRC aa 1620-1710 | TMEM145 ipTM 0.68; actin ipTM 0.51 in the Phase 3 handoff. | Reference computational design. |
| tail91 v2 | Cys11/52->Ser production candidate | TMEM145 median ipTM 0.55, sigma 0.010; actin median ipTM 0.51, sigma 0.011. | Developability-safe candidate for synthesis discussion. |
| RADA16-WH2 scaffold | Self-assembling SAP backbone with actin-binding handle | Top Phase 1 scaffold; later carried into PEPTIDE_TAIL91. | Assembly and multivalency engine. |
The central mathematical question is whether weak monovalent actin contacts become therapeutically useful when repeated along an assembled peptide scaffold. The current answer is yes under the Phase 4m calibrated floor, with direct binding assays still required.
| Run | Inputs | Output | Call |
|---|---|---|---|
| Phase 4j conservative | Kd_mono 10 mM, C_eff 50 mM, N=4 | Kd_eff 80 uM | Fail by 1.6x; forced the next proof. |
| Phase 4j nominal | Kd_mono 5 mM, C_eff 100 mM, N=5 | Kd_eff 31 nM | Pass, but too assumption-heavy to carry the claim alone. |
| Phase 4n Monte Carlo | Kd_mono 1-20 mM, C_eff 13.4-500 mM, N=3-6 | 80.3% ideal pass; 59.0% stressed pass | Identified Kd_mono as the bottleneck. |
| Phase 4m conservative rerun | Kd_mono 1.48 mM, C_eff 50 mM, N=3 | Kd_eff 1.30 uM | Pass; promotion from tentative-S to S. |
| Phase 4m higher valency | Kd_mono 1.48 mM, C_eff 50 mM, N=4 | Kd_eff 38.39 nM | Pass with margin if four contacts are physically available. |
Anti-handwave separation: which bricks are already in the literature, what h09 extrapolates from each one, and what residual risk remains. The unproven h09 bet is the integrated material bridge in the cochlear hair-bundle environment, not the individual bricks.
| Established precedent | h09 extrapolation | Remaining risk |
|---|---|---|
| STRC-related hearing loss is a human DFNB16 phenotype with mostly mild/moderate onset and a long residual-cell window. GeneReviews STRC / DFNB16 (2024) | h09 scopes only STRC-loss ears with residual OHC function, not broad hearing loss. | Late-stage disease or absent OHCs may be outside the response window. |
| Horizontal top connectors are mechanically load-bearing in mature OHC bundles. Strc-null/top-connector loss reduces bundle stiffness ~60% and damping ~74%. Verpy 2008; Verpy 2011; Rivera/Dulon 2019 | A synthetic connector-like material could be useful if it restores even partial inter-stereocilia cohesion. | Native STRC geometry and timed maturation may matter more than generic mechanical bridging. |
| WH2 domains bind G-actin with measured ITC Kd: WAVE2 52 nM, WIP 160 nM, MIM 230 nM, WASP 250 nM, Tbeta4 760 nM. Chereau 2005 PNAS, Fig. 1C | h09 uses a WH2-class actin-binding motif; Phase 4m places h09 in the WH2-family soft-floor band. | WH2 x F-actin side-binding Kd is not directly measured; Phase 4m is a soft floor, not a measurement. |
| Polyvalency/avidity can convert weak monovalent binding into useful effective binding when geometry is matched (Mammen / Jencks / Karpen formalism). Mammen, Choi, Whitesides 1998 Angew Chem | h09's RADA16 display is meant to turn weak mono WH2 contacts into cooperative effective binding. | Flexible or misregistered display can erase avidity; Phase 4k remained confidence-inconclusive on geometry. |
| RADA16/SAP scaffolds self-assemble into nanofibers and hydrogels; characterized fiber length 615 +/- 104 nm; 0.6-3 mM hydrogel formation range. Yokoi/Kinoshita/Zhang 2005 PNAS | h09 uses a low-fraction tail-bearing RADA16 blend, not pure 134-aa tail91 self-assembly. | Tail91 is far beyond characterized appendage length; rheology/self-assembly must be measured first. |
| Round-window entry into cochlear perilymph is experimentally quantifiable; primary RWM permeability and scala-tympani clearance parameters exist. Salt & Ma 2001 Hear Res | h09 delivery is a local exposure feasibility problem, not systemic delivery. | 14 kDa peptide/hydrogel permeation is extrapolated; target-zone localization remains unproved. |
| Independent Boltz-2 WH2 controls do not place h09 below sequence-destroyed negatives. Phase 4p, this work (2026-04-27) | The Phase 4m WH2-family claim survives one matched alternative-model sanity check. | Boltz does not recover the ITC rank order cleanly; supportive only, not affinity evidence. |
| h09's novelty is the combination: local peptide-material display + WH2/avidity + synthetic HTC-style mechanical rescue. h09 hub, this work | The reviewer question is whether this combination merits the minimum feasibility package. | The combination has no precedent; it must earn feasibility through localization, toxicity, and rescue assays. |
Hostile-read checklist run before external review. Each row pairs the strongest objection with the current answer, residual risk, and the smallest concrete check that could move the needle.
| Objection | Current answer | Residual risk | Minimum next check |
|---|---|---|---|
| AF3 is not a Kd predictor. | Agree. Phase 4m uses matched WH2-family rank calibration only, not absolute AF3 thermodynamics. | The rank relation could be a model artifact. | Run an orthogonal WH2 panel (positive references + h09 + scrambled + alanine) under one protocol. Phase 4p Boltz-2 already does the first pass. |
| WH2 x G-actin does not prove WH2 x F-actin. | Agree. Phase 4m uses Tbeta4 G-to-F scaling as an explicit soft floor. | Scaling may not transfer to h09 WH2 or to filament-side binding. | Label Kd_F = 1.48 mM as a soft floor, not a measurement. Reserve a filament assay for paper-grade C_eff / F-actin support. |
| Short WH2 motifs are weakly represented by structure models. | Phase 4k is filed as INCONCLUSIVE and h09 S does not rely on it. | A reviewer may discount all motif-model outputs. | Lead with primary WH2 Kd literature and Phase 4m calibration; use Phase 4k as method discipline, not support. |
| Hydrogel may not reach the hair-bundle target zone. | Delivery remains a feasibility premise, not a closed proof. | Ototopical / RWM / LIFU delivery could fail before biology matters. | Run the precedent-separated delivery panel: RWM permeability, LIFU permeation, peptide localization, hair-bundle imaging. |
| Self-assembly could cause local toxicity or occlusion. | Current dose window is computational and class-based only. | Local ototoxicity, vestibular exposure, or gel occlusion could kill the route. | Step 4 of the feasibility package: explant localization, dose titration, washout, toxicity before any rescue claim. |
| A synthetic bridge may not rescue STRC-null OHC mechanics. | h09 is a synthetic substitute concept, not STRC restoration. | Native STRC geometry, timing, or protein interactions may be required. | Define success as partial mechanical rescue. Ask reviewers which assay can falsify mechanical benefit fastest. |
| Patient window may be narrower than assumed. | Best fit is STRC-loss with residual outer hair cells and mild/moderate phenotype. | Late disease or advanced OHC loss may be unrecoverable. | Specify genotype, residual-cell inclusion criteria; avoid broad STRC-cure language. |
| Phase 4k did not independently pass. | Correct. Phase 4k is a caution: tandem-proxy geometric contacts yes, confidence-supported contacts no. | Reviewer may treat this as contradiction rather than model limitation. | State the hierarchy plainly: Phase 4m is gate-closing; Phase 4k is a negative-control-style limitation. |
| The hypothesis is a device/material concept, not a gene therapy. | Correct. h09 is local peptide-material therapy. | Wrong reviewer set will reject for the wrong reasons. | Send to biomaterials and inner-ear delivery reviewers before gene-therapy-only review. |
| Addressable population may be overstated. | Scope is STRC-loss with viable residual OHCs. | Impact claims could look inflated. | Use conservative prevalence and response-window language; do not imply universal STRC-null rescue. |
Phase 4p — independent model family asked the same WH2 question. h09 stays inside the positive-reference band and above all sequence-destroyed negatives. Boltz does not recover the Chereau ITC rank order cleanly (WIP false-low, Tbeta4 false-high), so this is supportive, not affinity evidence.
| Construct | Role | Boltz ipTM | Read |
|---|---|---|---|
| MIM | positive | 0.925 | high positive |
| Tbeta4 | weak reference | 0.914 | false-high vs ITC |
| WASP | positive | 0.873 | positive |
| h09 WH2 | query | 0.706 | in positive-reference band |
| WAVE2 | positive | 0.699 | positive, not rank-best |
| scrambled h09 WH2 | negative | 0.633 | below h09, close |
| alanine h09 WH2 | negative | 0.588 | below h09 |
| WIP | positive | 0.340 | false-low positive |
| poly-GS spacer | negative | 0.319 | low negative |
Five sequential gates. Do not run Step N unless Steps 1…N−1 pass. A clean failure at any gate is still useful: it gives a sharp falsifier and prevents money from being spent on a vague "maybe the model was right" story.
Can the h09 material be made cleanly enough to test?
tail91_v2, plain RADA16, scrambled control, alanine-WH2 control, fluorescent tracer aliquot.
HPLC purity (>=90% discovery, >=95% explant), LC-MS intact mass, analytical HPLC trace, endotoxin for explant lots, lyophilized solubility.
Identity / purity / solubility / endotoxin all clear; tail91_v2 + plain RADA16 in hand.
Peptide cannot be made cleanly or remains insoluble/aggregated under assay buffer.
Do low-fraction tail91_v2 / RADA16 blends preserve beta-sheet fiber and gel behavior?
0%, 1%, 5%, 10%, 100% tail91_v2 in plain RADA16; matched buffer / pH / salt / temp.
CD spectroscopy (216-218 nm beta-sheet), AFM/TEM fiber morphology, DLS aggregate distribution, rheology G′ > G″, visual handling.
5% tail91_v2 blend forms RADA16-like beta-sheet/fiber/gel material without large-aggregate failure.
Long-tail blend destroys assembly or forms micron-scale aggregates.
Does h09 separate from sequence-negative controls on actin and TMEM145 axes?
h09 tail91_v2 blend (Step 2 pass condition), plain RADA16, scrambled control, alanine-WH2, poly-GS, optional WH2 positive reference.
A) G-actin: fluorescence anisotropy / MST / BLI / pull-down. B) F-actin: co-sedimentation, TIRF bundle imaging. C) TMEM145: best available extracellular/GOLD proxy. Dose curves with all controls on the same axes.
Dose-dependent h09 signal separates from scrambled / alanine / poly-GS controls on each axis.
Negatives match h09, or no F-actin / TMEM145 signal is separable.
Does the material reach the hair-bundle region without obvious local harm?
Organ of Corti / cochlear explants. Exposures 0.3, 1, 3, 10 uM peptide-equivalent + vehicle + plain RADA16 controls.
Hair-bundle localization (fluorescent tracer aliquot + unlabeled toxicity controls), washout/reversibility, OHC survival (myo7a / prestin / phalloidin), bundle morphology, MET / FM dye / Ca proxy.
Bundle-region enrichment at non-toxic doses, no irreversible occlusion, no OHC loss vs controls.
No bundle-zone localization, OHC toxicity, irreversible occlusion, or vehicle-independent clumping.
Does h09 produce a measurable mechanical / functional benefit?
Vehicle, plain RADA16, scrambled/alanine control, h09 at the best Step 4 non-toxic/localizing dose, optional positive rescue if available.
Hair-bundle stiffness / deflection, MET-current or FM-dye proxy, high-speed bundle motion coherence, Strc-null vs WT comparison if available. Predefine the primary endpoint before starting.
h09 improves the prespecified primary endpoint vs vehicle and sequence-negative controls without toxicity.
No improvement, or apparent improvement is paired with toxicity/occlusion.
If Steps 1–4 pass, h09 becomes a legitimate pre-rescue feasibility candidate. If Step 5 also passes, h09 becomes a real experimental rescue hypothesis rather than a computational feasibility concept.
H09 is ready to present as a computationally promoted, reviewer-packaged hypothesis for a first synthesis and binding campaign. It is not yet a validated drug, material, or hearing rescue result.
Read the page as the strongest current case for starting wetlab work, with Phase 4m as the key promotion evidence and Phase 4k as the main geometry caution. Reviewer-pack prose cannot move ranking — only new computational or experimental evidence can change h09 tier, mech, deliv, or misha_fit.