STRC h09 Phase 4e_v2 Blend Scaffold

Closes 1 of 3 A→S promotion gates for STRC Synthetic Peptide Hydrogel HTC flagged by STRC h09 Parameter Provenance Audit 2026-04-23. Specifically: the 118 aa tail91 > 12 aa empirical RADA16 modification limit gap.

Verdict: 5 % molar blend (plain RADA16 : RADA16-tail91 = 19:1) is literature-defensible and yields ~2000× excess valency vs putative stereocilia TMEM145 site density. Valency is NOT the limiting factor at the recommended blend.

Problem restated

Phase 3b validated the tail91 construct in AF3 (ipTM 0.68 for both TMEM145 and actin, full-construct 118 aa design). BUT per Frontiers 2021 review (PMC8216384):

  • ≤12 residues: β-sheet scaffold assembly preserved
  • 12–50 residues: monotonic viscoelasticity decline
  • >50 residues (esp. globular): untested in published lit
  • ALK-tag (+3 hydrophobic): rapid collapse to spheres

Our tail91 = 91 aa = ~10× beyond characterized safe limit. Phase 4d/4e dose calculations that assume 100% tail91 monomers assemble into functional scaffold are not defensible.

Standard workaround (Zhang / Gelain group, RGD-osteogenic precedents): molar blend of plain RADA16 + RADA16-tail-fusion. Plain scaffold carries β-sheet; tail projects out, displaying binding motif.

Model

Script: hydrogel_phase4e_v2_blend_scaffold.py. Literature-anchored parameters (from literature-params/rada16-geometry.md):

ParameterValueSource
RADA16 monomers/nm fibril4.26Paravastu 2014 + Cormier 2012
Mean fibril length615 nmYokoi 2005
Fibril cross-section3×2.5 nm tapeParavastu 2014
Fibers per bundle3–7 (mean 5)Zhang 1993
Empirical tail limit (preserved)≤12 aaFrontiers 2021
Empirical tail limit (monotonic)12–50 aaFrontiers 2021
Stereocilia spacing9 nmKrey 2016
Perilymph vol34 μLDhanasingh 2021

Sweeps: φ_tail ∈ {0.01, 0.05, 0.10, 0.20, 0.33, 0.50} × gel [%w/v] ∈ {1.0, 2.0}.

Results (φ_tail × gel grid)

φ_tailgel %flagtails/fibriltails/bundle shellvalency ratio (low est)dose vs 100% tail91
0.011.0%PASS267.8×10⁶261×15.5×
0.012.0%PASS261.6×10⁷522×15.5×
0.051.0%PASS1313.2×10⁷1069×3.8×
0.052.0%PASS1316.4×10⁷2138×3.8×
0.101.0%WARN2625.2×10⁷1743×2.3×
0.102.0%WARN2621.0×10⁸3487×2.3×
0.201.0%RISK5247.6×10⁷2547×1.6×
0.202.0%RISK5241.5×10⁸5094×1.6×
0.33RISK8651.3×
0.50FAIL13101.1×

Valency ratio = (tails in gel shell per OHC bundle) / (putative TMEM145 sites per bundle, low estimate 30×10³).

Key finding: valency is NOT the binding constraint

At φ_tail = 0.05 (PASS) + 2% w/v gel: 2000× excess tail91 per bundle over conservative TMEM145 site density. Even at 0.01 (15.5× dose cost vs 100% tail91), 261× excess. The scaffold carries so many tail monomers per bundle that even stringent phase-separation loss (e.g., 90% of tails clustered away from the interface) leaves ~200× surplus.

This reframes the engineering problem: optimize assembly integrity first, valency second. Pick the lowest φ_tail that still assembles robustly.

Recommendation

Design point: φ_tail = 0.05, gel 2% w/v.

Justification:

  • Within Gelain osteogenic hybrid range (literature precedent, though with shorter tails).
  • β-sheet scaffold assembly dominated by 95% plain RADA16 — minimal disruption.
  • Dose cost: 3.8× total peptide mass vs hypothetical 100% tail91. At Phase 4e 0.13–1.32 mg therapeutic window, that shifts to 0.5–5.0 mg blend. Still well within cochlear injection tolerance (typical intra-tympanic: 1–10 mg).
  • Valency ratio ~2000× TMEM145 sites (low estimate) — comfortable over-provisioning even under phase separation or clustering.

Fallback for risk tolerance:

  • φ_tail = 0.10 (WARN): 2.3× dose cost, retains Gelain-range-adjacent feasibility. Still 3500× valency ratio. Use if experimental data from wet-lab rheology shows 5% tail91 insufficient for binding.

What this DOES NOT close

GateStatus
Gate 1: 118 aa tail vs 12 aa limitCLOSED by 5% blend extrapolation
Gate 2: WH2×F-actin side-binding Kd unmeasured⚠ Still open (Phase 2c wet-lab bundling assay needed)
Gate 3: STRC×TMEM145 Kd unmeasured⚠ Still open (SPR/BLI wet-lab OR Path B AF3 ipTM→Kd calibration [task #14])

The blend model is lit-backed extrapolation, not primary data. The true test is rheology + β-sheet CD spectrum on blended peptide at 5% φ_tail — wet-lab experiment, not in scope here.

Caveats

  • TMEM145 site density is order estimate. “30 × 10³ sites per bundle (low)” = 1000/stereocilium, typical membrane-receptor order. Not measured in primary lit. Valency ratios depend on this; factor 10× error in estimate still leaves 200× surplus at 5% φ_tail.
  • Frontiers 2021 characterizes pure constructs (0% plain RADA16), not blends. Extrapolating “assembly behavior at 5% tail” from “100% tail of length X” is a model, not a measurement. Gelain group hybrids validate the principle for shorter tails; no peer-reviewed blend data exists for >50 aa tail + ≥5% loading.
  • Phase separation risk not modeled. Monte-Carlo or coarse-grained MD would bound the tendency of tail91 monomers to self-segregate. Would require separate simulation.
  • Proteolysis/clearance unchanged. This model only addresses scaffold geometry + valency, not PKPD.

Ranking delta

  • #9 Hydrogel HTC: A held (tier unchanged). Mech 4 held. Deliv 3 held.
  • Triple S-tier gate: 1/3 closed; 2/3 remain. Cannot promote to S this turn.
  • Next step for h09 updated: “4b AF3 submit + Phase 2c WH2 assay + TMEM145 SPR” (was “4b + blend model + lit-backed re-run”; blend is now done).

Connections