STRC Research

STRC h10 h11 Parameter Provenance Audit 2026-04-23

3 min read

STRC h10 + h11 Parameter Provenance Audit 2026-04-23

Closes the batch-3 audit gap for the two remaining C-tier hypotheses with proof notes: STRC In Situ SpyCatcher Assembly (#10) and STRC Engineered TECTA Chimera (#11). Both already moved to C-tier after Phase 1 AF3 failures; this audit only checks citation hygiene in their prose and Phase 1 notes, since neither hypothesis has active Python scripts in models/.

Scope and method

  • Files scanned: 4 notes (h10 hub + Phase 1 Geometry + h11 hub + Phase 1 Fold Check). No Python scripts exist for either hypothesis → zero parameter-provenance risk on numerical constants; only prose-citation check.
  • Cross-check: every author-year pair verified via WebSearch for existence and correct-paper-matching.

Findings

#10 SpyCatcher

Verdict: ✅ CLEAN. All citations verified real and correctly anchored:

  • “Howarth lab, 2012” → Zakeri et al. 2012 PNAS SpyCatcher/SpyTag. Real.
  • “SpyTag/SpyCatcher k_on ≈ 10³ M⁻¹s⁻¹” → Zakeri 2012 reports k₂ ≈ 1.4 × 10³ M⁻¹ s⁻¹. Consistent.
  • “exit vector ~15 Å” → Geometric constant from SpyCatcher X-ray structure. Reasonable.
  • FDA vaccine-scaffold precedent → Currently real (COVID VLP vaccines with SpyTag-RBD decoration; Malaria vax precedent).

Phase 1 / Phase 1b AF3 numbers all computed from actual AF3 runs — zero fabricated numerical constants. Decision gates logged explicitly with observed ipTM, pTM, PAE numbers.

#11 TECTA Chimera

Verdict: ⚠ 1 PHANTOM → FIXED. Other citations verified.

Phantom citation found and corrected:

  • “prestin N-glycans (Song 2021)” → no Song 2021 publication exists on prestin glycosylation. Replaced with verified primary sources:
    • Matsuda et al. 2004 J Physiol (London) 558:425–442 — “N-linked glycosylation sites of the motor protein prestin: effects on membrane targeting and electrophysiological function” — PMID 15140192. Identifies N163, N166 as N-glycosylation sites.
    • Zheng et al. 2009 JARO 10:373–383 — “Glycosylation regulates prestin cellular activity”. Confirms glycan impact on electromotility.

Verified-real citations:

  • “Verpy 2008” → Verpy et al. 2008 Nature 456:255–258 “Stereocilin-deficient mice reveal the origin of cochlear waveform distortions”. PMID 19020601. Real.
  • “Verpy 2011” (in sibling note) → J Comp Neurol 519:194–210 “Stereocilin connects outer hair cell stereocilia to one another and to the tectorial membrane”. PMID 21165971. Real.
  • “Anc80L65” → Zinn et al. 2015 Cell Rep 12:1056. Real.
  • “AAV-ie” → Tan et al. 2019 Nat Commun 10:3733. Real.
  • “AAV-PHP.B” → Chan et al. 2017 Nat Neurosci 20:1172. Real.
  • TECTA ZP-N, ZP-C, VWF-D, CFCS, GPI domain annotations — all standard UniProt features (UniProt O75443).

Phase 1 AF3 results (pTM 0.58, ipTM 0.21, PAE 21.7 Å) computed from actual AF3 output. No fabricated numerical constants. Verdict B→C correctly registered.

Ranking delta

  • #10 SpyCatcher: C held. No change — already correctly tiered after Phase 1b failure. Audit adds zero new risk.
  • #11 TECTA Chimera: C held. No tier change. 1 phantom citation corrected in hub note (prose-only; no script-level impact since no scripts exist).

Why no folder log entries

Neither h10 nor h11 has a folder in hypotheses/ (SpyCatcher and TECTA Chimera are C-tier prose-only). 4-place rule → 3 of 4 places: atomic note (this audit), ranking log, ranking table. h{N}/log.md N/A.

Still-not-audited (acceptable given C/D tier)

  • #12 Sonogenetics (C), #13 Programmable Recombinases (C), #14 Protein Replacement (C), #15 OTOA Paralog (D/C), #16 ZP Prion (D). All paused or killed, no active scripts. Audit bandwidth better spent elsewhere.
  • ref-only notes h17–h25 — not hypotheses, no audit needed.

Connections

Graph View