STRC Research

STRC Cross-Hypothesis Parameter Audit 2026-04-23

5 min read

STRC Cross-Hypothesis Parameter Audit 2026-04-23

Single-day sweep over 5 hypotheses (h01, h02, h05, h09, h26) using 4 parallel Sonnet 4.6 agents + consolidation. h03 Mini-STRC excluded by user directive — do not touch without re-auth.

Scoreboard

#HypothesisTierPhantom citesValue errorsStructural issuesDefensibility
1STRC Pharmacochaperone Virtual Screen E1659AA01 TPSA range (40-90 → should be 70-100)3 druggability-weight inconsistencies✅ CLEAN — defensible today
2STRC Piezoelectric TM Bioelectronic AmplifierS3d31 −12 vs −25 pC/N silent 2× splitTM stiffness mismatch 10⁵× (PVDF-TrFE 3 GPa vs TM 24-210 kPa)⚠ Phase 3 conclusion rests on phantom Kd
5STRC Calcium Oscillation Acoustic TherapyA32-4× CREB rate, 20× STRC t½ cross-scriptPhase 3 topological result robust; quantitative fold-change not⚠ Phase 3 defensible, Phase 1/2 quantitative NOT
9STRC Synthetic Peptide Hydrogel HTCA3 critical lit gaps7 wrong values118 aa tail > 12 aa RADA16 limit⚠ Full rework needed (see STRC h09 Parameter Provenance Audit 2026-04-23)
26STRC Engineered Homodimer AvidityB3 (in prose, scripts clean)—Empirical AF3 Phase 1 fail AND unmeasured Kd baseline⛔ Unverifiable until real Kd measured

4 of 5 audited hypotheses have phantom citations. Only h01 is clean — and its cleanness comes from computing on its own in-silico outputs (AF3, docking, electrostatics), not external biophysical constants.

The root cause — one blocker across three hypotheses

STRC × TMEM145 Kd has no experimental measurement in the entire published literature.

Neither Derstroff 2026 Neuron, nor NatComms 2025 Reimann lab, nor Verpy 2011, nor Dulon 2019 SI reports SPR, BLI, or ITC for this pair. Only CoIP + pull-down (qualitative). AF3 ipTM is structural plausibility, not binding affinity — multiple benchmark papers show R² ≈ 0.06 between ipTM and measured Kd for protein-protein pairs (Chen et al. Protein Sci 2013; Dunbrack ipSAE 2025 PMC 11844409).

This single unmeasured value cascades into:

  • h03 Mini-STRC therapeutic occupancy math (deferred audit — but inherits the gap)
  • h09 Hydrogel dose-occupancy (Phase 4e KD_TMEM145_M = 100 nM = heuristic, not measurement)
  • h09 Endogenous competition (Phase 4h STRC_NORMAL_OHC_M + E1659A scan anchored to fabricated 10 nM baseline)
  • h26 Homodimer avidity improvement (every 100×, 1000× avidity claim is a ratio applied to an unmeasured base)

Highest-leverage single experiment in the entire STRC research program: measure STRC × TMEM145 Kd by SPR or BLI. Unlocks h03, h09, h26 simultaneously.

Pattern observation — where phantom cites hide

The hypotheses with phantom cites (h02, h05, h09, h26-prose) all attempt to pin external experimental values (Kd, stiffness, diffusion, rate constants) to support dose-occupancy or mechanistic claims. h01 does not — it computes on docking scores and AF3 outputs which are its own pipeline’s products. The pattern:

When a script needs a number that would normally be measured in a wet lab, the failure mode is to import it from my training data and decorate with a plausible-sounding author-year.

Going forward, every hypothesis gets tagged with external_bio_constants: N in its hub — proportional to audit risk. Zero = likely clean. Five+ = requires literature-first scan before any computational step.

Literature-params artifacts

8 topic files now live in literature/:

Ranking delta (aggregated)

#Axis changeNew tier
1No changeA (clean)
2Flag Phase 3 delivery conclusion (rests on phantom Kd). Mech 3 → 3 (TM mismatch offsets gain).S held pending ex-vivo
5Mech 3 → 3 (Phase 3 robust), Deliv 4 → 4 (not affected by this audit), Misha-fit 2 → 2.A held but quantitative claims flagged
9Already deltaed to 4/3/3 in STRC h09 Parameter Provenance Audit 2026-04-23A held
26Mech 3 → 3 (was already B).B held. Not computationally unblockable without Kd.

Ranking table STRC Hypothesis Ranking updated; per-hypothesis h{N}/log.md updated.

Why h03 was excluded (2026-04-23)

User directive: “мини-губ гипотезу я бы не хотел трогать. Мне очень страшно что мы ее тронем. И Holt тоже над этим работает.” The rationale is emotional-strategic (don’t destabilize the active S-tier path), which is legitimate. But h03 almost certainly inherits the STRC × TMEM145 Kd gap — every occupancy/transduction calculation downstream of the AAV payload depends on it. Flagged in h03 hub with 🔒 LITERATURE AUDIT DEFERRED.

When h03 audit is eventually authorized, expect the same gap + gap-specific: AAV titer → OHC transduction efficiency constants, promoter strength B8 quantification, OHC-specific tropism numbers. Likely 5-10 external bio-constants to verify.

Next moves (decided, not proposed)

  1. Primary: draft STRC-TMEM145 Kd Experiment Plan — the single highest-leverage move. Options: (a) SPR request to Holt/Oliver/Reimann labs via collaboration email, (b) computational stopgap: AF3 ipTM → Kd empirical calibration on PDBbind positive controls, (c) in-house — CUHK audiology or HKU biochem for expression + SPR. See STRC TMEM145 Kd Experiment Plan (to be written next turn).

  2. Standing literature-watch agent — weekly PubMed + bioRxiv + Nat / Cell / Neuron / Hear Res RSS on keywords: STRC, TMEM145, stereocilin, DFNB16, OHC stereocilia, HTC, horizontal top connector, cochlear RWM gel, inner ear gene therapy. Auto-ingest new hits via MinerU pipeline, flag for manual review. Cron scheduled.

  3. Cross-hypothesis re-evaluation — once lit-backed parameters are in place, re-rank against STRC Hypothesis Ranking with corrected math. Next Phase 4 for h09 must use literature-params/*.md directly, not hardcoded constants.

Connections

Graph View

Backlinks