STRC Paper Draft Outline

First-draft nucleation for the h01 pharmacochaperone story. Voice = Egor’s (terse, mechanistic, no hype). Target venue: Nature Chem Biol / PNAS (mechanism-heavy) or J Med Chem (medchem-heavy), depending on wet-lab completion. This is an outline, not prose — fill from phase proofs when writing.

Working title candidates

“An electrostatic pharmacochaperone mechanism rescues the DFNB16-associated E1659A variant of stereocilin”
“Pocket-electrostatic pharmacochaperone design for misfolded stereocilin: computational framework and first-in-class hit”
“How a missense mutation creates a drug pocket: ensemble electrostatics guide pharmacochaperone discovery for DFNB16 hearing loss”

One-paragraph abstract (target ~200 words)

Pharmacochaperones rescue misfolded proteins by binding and stabilizing them. For the autosomal recessive hearing loss DFNB16 (OMIM 603720), the c.4976A>C variant produces E1659A stereocilin which fails to fold properly. Matched-ensemble Poisson-Boltzmann electrostatics on 20 ns each of WT and E1659A full-length molecular dynamics (2 ns per state, AMBER14SB/TIP3P) shows the mutant K1141 pocket is +4.53 ± 0.46 kT/e more positive than wildtype (Welch t=9.81, p = 6.9 × 10⁻¹², Cohen’s d = 3.10; −2.79 ± 0.28 kcal/mol favourable for −1e anions). Coarse Coulomb scoring in AutoDock Vina misses this asymmetry because PDBQT Gasteiger charges net to zero on acidic ligands, creating a +0.5 kcal/mol WT-bias artefact. By integrating ensemble APBS scoring alongside Vina docking (combined objective Vina_ΔG + ΔG_formal_ensemble), we identify 2-amino-quinoline-3-carboxylic-acid scaffolds with acidic-warhead placement matched to the pocket’s mean potential. Systematic structural exploration across 444 candidate analogs (60 v5 + 384 v5.2) yields a 14-compound shortlist with combined-score < −7.0 kcal/mol AND 10-gate ADMET-clean (15/30 v5.2 clean; naphthyl-containing 1/11 clean), dominated by non-planar cage tails (adamantyl, 1-indanyl). The top combined lead v5.2__aq3__adamantyl__CONHOH__-Cl (combined −8.03 kcal/mol, logP 1.94) is the first in-silico clean pharmacochaperone candidate for stereocilin E1659A.

1. Introduction (2-3 paragraphs, ~500 words)

§1.1 DFNB16 and stereocilin

DFNB16 biology: hair-cell tectorial-membrane–stereocilia attachment; STRC on 15q15.3 encodes 1775-aa extracellular protein with three LRR domains
Incidence: ~7-10% of autosomal recessive nonsyndromic hearing loss in caucasian populations; highly prevalent in conductive-hearing-loss diagnostic panels
Compound heterozygosity common; large deletions + missense variants account for most cases
Misha’s patient: c.4976A>C / c.[deletion], maternal E1659A allele → entry point for monotherapy or combo-therapy strategy
Previous therapeutic attempts: none FDA-approved; current clinical recourse is hearing aids
Cite: Misha Compound-Het Therapy Stack Model, clinical genetics reviews

§1.2 Pharmacochaperones as precedent

Tafamidis/TTR: FDA-approved 2019 pharmacochaperone rescues transthyretin tetramer from dissociation → hereditary amyloidosis
Cystic-fibrosis VX-809 / ivacaftor: correctors stabilize ΔF508-CFTR folding
Generic pharmacochaperone paradigm: compound binds a partially-folded state, lowers the kinetic barrier to native state, rescues function
Key design requirement: druggable pocket on the misfolded form that is absent or unfavourable on the WT form

§1.3 The electrostatic asymmetry hypothesis

E1659A removes a −1e Glu sidechain in the vicinity of a K1141-centred pocket
Net Coulomb effect on the pocket interior: become more positive by ΔΦ ≈ +e/(4πε₀εp · r) · exp(−r/κ⁻¹) ≈ several kT/e for r ~ 15-20 Å
Prediction: acidic ligands should bind the mutant pocket with Coulomb preference, without needing a specific geometric salt-bridge anchor
This work: computationally validates the asymmetry, demonstrates a Vina limitation that hides it, designs a first-generation Coulomb-aware medchem campaign

2. Results

§2.1 The E1659A pocket is +7 kT/e more positive than WT at AF3-level structure

Source: STRC h01 Phase 5j APBS WT vs Mutant Pocket Electrostatics 2026-04-24
APBS nonlinear PBE on chain A AF3 structures (PARSE force field, pH 7.4, 150 mM NaCl, ε_p=2, ε_s=78.54, 310 K)
WT pocket mean φ = −1.60 kT/e (52% positive); mutant = +5.50 kT/e (73% positive); Δ = +7.10 kT/e (+4.37 kcal/mol for q=−1e anion)
Pocket-interior Debye-screened Coulomb interpretation (Eq. 1)
Figure 1: WT vs mutant pocket isosurface, −1 to +10 kT/e; shows the positive-fraction transition

§2.2 Vina docking is blind to this asymmetry

Source: STRC h01 Phase 4c-v3b WT Decoy on v3b YELLOW 2026-04-24 + STRC h01 Phase 4d K1141A Double-Mutant Decoy 2026-04-24
Phase 4c-v3b on 29 v3b YELLOW: 0/29 ligands prefer mutant; median WT-bias +0.55 kcal/mol
Phase 4d K1141A decoy: K1141 removal costs only 0.18 kcal/mol → salt-bridge mechanism falsified
Gasteiger PDBQT ligand total q = 0 ± 0.004 e across all 29 v3b → Vina literally does not represent the formal COO⁻/tetrazolide anion
Figure 2: Vina ΔG distribution WT vs mutant, with zero-charge ligand schematic

§2.3 Pose-transplant APBS shows the rescue at per-pose resolution

Source: STRC h01 Phase 4i APBS Pose-Transplant Rescore 2026-04-24
Pocket-local Kabsch alignment (residues 1126-1156 Cα, RMSD 0.21 Å after alignment) enables cross-frame pose transplantation
For the same ligand geometry, 20/29 v3b prefer mutant pocket by APBS (median Δφ = +0.70 kT/e at acid centroid; formal-anion ΔG = −0.43 kcal/mol)
The remaining 9 ligands marginal — not failures, just pose-location sensitivity
Figure 3: per-ligand Δ_Vina vs Δ_APBS scatter; ~1 kcal/mol gap consistent with missing formal charge

§2.4 Ensemble MD confirms robustness under dynamics

Source: STRC h01 Phase 5k Ensemble APBS on Phase 5d Mutant MD 2026-04-24
2 ns E1659A full-length AMBER14SB/TIP3P MD (Phase 5d) → 20 snapshots
APBS re-run from scratch on each snapshot with self-consistent K1141-centred box: pocket φ = +5.99 ± 1.37 kT/e, 20/20 positive, range [+2.77, +8.70]
Phase 5j static +5.50 sits at ensemble median — not pathological frame
Phase 5k-B pose × ensemble: 400 pose-snap measurements, 92% positive, median formal-anion ΔG = −0.67 kcal/mol
Figure 4: pocket-φ across 20 MD frames; histogram of per-pose Δφ; ensemble-Coulomb re-ranking of v3b (Vina top-1 is only #3 by Coulomb)
(Pending: §2.4.x — Phase 5d-WT MD matched ensemble, will confirm the ~7 kT/e shift is a mutant-specific phenotype rather than a buffer effect; closes cryptic-pocket specificity claim jointly with Phase 5c-mutant)

§2.5 Coulomb-aware medchem redesign produces ADMET-clean leads

Source: STRC h01 Phase 8 v5 Library Coulomb-Aware Design 2026-04-24 + STRC h01 Phase 8c v5 ADMET-AI Triage 2026-04-24 + Phase 8d v5.2 (pending)
v5 library design axes: non-planar cage tails (adamantyl, BCO, norbornyl) + acidic-head bioisosteres (CONHOH, tetrazole, COOH, phosphonate) + R3 palette
Combined objective: combined_score = Vina_ΔG + α·ΔG_formal_ensemble with α=1.0
v5 v1 top combined: adamantyl__CONHOH__-Cl at −8.19 kcal/mol; logP 1.94; 95% ensemble-positive
v5 top-20 ADMET panel: 12/20 flag-free vs 0/10 v3b benchmarks — naphthyl tails fail DILI 93-95 pct; adamantyl/BCO/norbornyl are universally clean
Figure 5: v5 vs v3b combined score distribution; ADMET-gate pass fractions; top-5 lead structures

§2.6 Off-target selectivity (pending Phase 8f)

Cochlear-target panel: hERG, KCNQ4, Cx50, TMEM16A, TRPM4, BK
COX-1/COX-2 for NSAID-class historical context
Expected: v5.2 adamantyl series shows >100× selectivity vs hERG relative to v3b
Gate for wet-lab progression

3. Discussion

§3.1 Why Vina missed this for a decade

PDBQT Gasteiger is the community standard; neutral-netting behaviour is universal
Comparable blind-spots likely exist for any target with pocket-scale Coulomb asymmetry
Community should adopt ensemble-APBS rescoring as a standard second-pass filter for charged-ligand targets

§3.2 Design principle: pocket-Coulomb as a first-class medchem objective

Traditional: Vina ΔG → ADMET filter → wet-lab
This work: Vina ΔG + ensemble-APBS ΔG → ADMET filter → wet-lab
The Coulomb layer redirects chemistry toward mechanistically-mat hed scaffolds

§3.3 Implications for DFNB16 and related mutation-opens-pocket diseases

Similar mechanism possible for other STRC missense (R1581F, others — Phase 6 outlook)
General category: “loss-of-charge mutations” in buried regions may unintentionally create druggable pockets
Cochlear drug-delivery considerations: intracochlear route (Phase 6c hERG mitigation) vs systemic (now re-opened via ADMET-clean candidates)

§3.4 Limitations

Static protein + implicit solvent → next: holo MD + MM-PBSA decomposition
In-silico ADMET only → next: wet-lab patch-clamp + Caco-2 + hepatocyte clearance
Single mutation E1659A → next: generalize to R1581F, L1516V, D1632N

4. Methods (~2 pages in final paper)

AF3 structure prediction: inputs, seeds, confidence, limitations
MD protocol: AMBER14SB/TIP3P, OpenMM Metal-OpenCL, 2 ns production
APBS protocol: PARSE force field, nonlinear PBE, mg-auto mesh, box parameters
Vina 1.2.7: Meeko prep, exh=16-32, num_modes=3
Pocket-local Kabsch alignment: residues 1126-1156, RMSD 0.21 Å validation
ADMET-AI: DrugBank-approved-percentile gate at 90
Combined-score methodology: α=1.0 rationale, sensitivity analysis
All scripts in supplementary: STRC Computational Scripts Inventory

4.5 Update 2026-04-26 — figures rendered, methodological lemmas formalised

Today’s session (2026-04-26) materialised two of the previously-stub figures and added two paper-grade methodological lessons:

Figure 2 (rendered): ~/STRC/hypotheses/h01-pharmacochaperone/artifacts/phase5e_v2/paper_figure2.{png,svg} (123/110 KB). 3-panel composition:

Panel A: v5.2 phosphonate × Gasteiger PDBQT — perfect anti-correlation ρ ≈ −1 (Vina vs τRAMD)
Panel B: v5.2 phosphonate × dimorphite+meeko formal-charge prep — ρ persists at −0.7 (rigid-receptor artifact remains after charge fix)
Panel C: v5.3 sulfonamide × formal-charge — Vina vs Boltz-2 ipTM AGREE on rank-1, ρ flips positive
Title encodes the chemistry-class rule: anion ≥ −1.5 → AGREE | ≤ −2 → INVERT

Figure 3 (rendered): ~/STRC/hypotheses/h01-pharmacochaperone/artifacts/phase5m_trpm4/paper_figure3.{png,svg} (103/99 KB). 3-panel composition:

Panel A: STRC τRAMD × 5 v5.2 candidates (median bars + per-replica scatter, within-STRC spread 1.7×)
Panel B: TRPM4 τRAMD × 2 candidates (LEAD + mech-anchor) with replica_02 outlier visible at 53.4 ps
Panel C: τ_TRPM4 / τ_STRC ratio bars + Phase 5p gate horizontal line at 5× (red dashed) + parity line at 1× — both candidates 1.08–1.52× ≪ gate

Two methodological lemmas formalised as Brain notes (paper-grade, reusable):

Vina Method-Class Rule — Anion Charge Magnitude vs Pocket Geometry — Vina rigid is reliable for anion ≥ −1.5 e + linear-volume head; unreliable for anion ≤ −2 e + tetrahedral-volume head. Backed by Phase 5e-v2 (failure on phosphonate) and Phase 5q (success on sulfonamide). Operational pre-screen: dimorphite SMILES → check anion magnitude → decide method.
Within-Target to Cross-Target Spread Lemma — Kinetic-Selectivity Pre-Screen — within-target τRAMD spread σ × √n_replicas × scaffold-factor (1–2×) predicts cross-target ratio. Validated on STRC×TRPM4 v5.2 (within 1.7× → predicted ≤ 2.4× → measured 1.08–1.52× ✓). Saves ~5h cross-target compute when within-target σ is too small to clear the kinetic-selectivity bar.

Methodological position update: the paper’s previous “Vina misses pocket asymmetry due to Gasteiger zero-charge” claim (Section 2 / current Figure 2 stub) is now sharpened: there are TWO stacked artifacts, not one. Gasteiger is fixable (dimorphite+meeko, verified 2026-04-26); rigid-receptor remains and is chemistry-class-conditional. Paper Section 2 should be expanded to a “double-artifact decomposition” subsection. The chemistry-class rule is the falsifiable formulation of the artifact’s domain of applicability.

Kinetic-selectivity claim (Section 2 / current Figure 3 stub): previously “kinetic discrimination requires same-class head-group spread > 5×“. Now empirically falsified at both within-target (Phase 5m, 1.7×) AND cross-target (Phase 5m TRPM4, 1.08–1.52×) levels for the v5.2 head-class. Paper should re-position from “we measured it and got 1.7×” to “we measured it AND derived a pre-screen lemma that predicts cross-target ratio from within-target spread within 2×“.

Open: chemistry-class boundary of the lemma. Validated only on v5.2 acyl-amide / phosphonate adamantyl-derived. Phase 5q v5.3 acyl-sulfonamide TRPM4 τRAMD (next session) is the second validation point. Scripts ready: phase5q_strc_dock_v53.sh, phase5q_strc_taramd_v53.py (within-target), then decision gate on σ_v53, then conditionally phase5q_trpm4_dock_v53.sh + phase5q_trpm4_taramd_v53.py.

4.6 Update 2026-04-26 (evening) — v5.3 sulfonamide class delivered + lemma second validation point closed

Phase 5q STRC within-target τRAMD on v5.3 acyl-sulfonamide PASS-list (3 lig × n=20 rep, paper-grade): 60/60 UNBOUND in 10.8–59.8 ps biased MD (~50 min wall on M5 Max CPU OpenMM after ~14 min n=5 pilot). σ_within (max/min mean τ̄) = 1.358× (n=5 was 1.45×, tighter at n=20); σ_within (median-based, outlier-robust) = 1.167×. Rank order at n=20: adamantyl_-CF3 (21.59±2.50) > adamantyl_-Cl (17.46±0.60) > 1-indanyl_-Cl (15.90±0.82); all three statistically tied by median (14.95–17.45 ps band). → STRC h01 Phase 5q v5.3 STRC Within-Target tauRAMD 2026-04-26.

Lemma second validation point — pre-emptive cross-target falsification at n=20: σ_within = 1.36× (mean) / 1.17× (median) → predicted STRC:TRPM4 ≤ 2 × σ_within = 2.72× / 2.33×, well below required 5× kinetic-selectivity gate. By the Within-Target to Cross-Target Spread Lemma — Kinetic-Selectivity Pre-Screen the v5.3 sulfonamide class is falsified pre-emptively on STRC alone with n=20 statistical robustness; fresh TRPM4 v5.3 dock + τRAMD compute is not warranted (~3 h saved). The lemma’s 2× scaffold-factor was conservative on v5.2 (predicted ≤ 2.4×, observed 1.08–1.52×); on v5.3 at n=20 it predicts ≤ 2.7× and the class verdict is binding. Affinity-rank-1 ≠ τ-rank-1 at n=20: three independent affinity proxies (Boltz-2 dynamic, Vina rigid n=1, Vina ensemble n=20 mut-snaps) all rank 1-indanyl_-Cl first; τRAMD ranks it last by mean and median. The split is no longer a small-n artifact — it is a real method-class divergence between equilibrium-pose-affinity vs kinetic-residence axes.

Figure 4 (rendered): ~/STRC/hypotheses/h01-pharmacochaperone/artifacts/phase5q_strc_v53/paper_figure4.{png,svg}. 2-panel:

Panel A: v5.3 STRC τRAMD per-ligand (mean ± SEM bars + per-replica scatter; within-target σ annotated)
Panel B: lemma cross-target bound (2.9×) vs required gate (5×) with σ_within ≥ 3× threshold for “run TRPM4” overlaid

Phase 5e v5.3 mut-ensemble re-dock (n=20 Phase 5d snaps): rank-1 by mean and median = 1-indanyl_acylsulfonamide_SO2Me_-Cl (mean −0.67 ± 3.10, median −1.39, best −4.59 kcal/mol). Adamantyl scaffolds wide-spread (std ≈ 11) — best-pose ≤ −4.6 but high-positive median → relies on snapshot-cherry-picking, not a robust mut-ensemble binder. → STRC h01 Phase 5e v5.3 Mut Ensemble Dock 2026-04-26.

Phase 8c v5.3 ADMET-AI triage (10-endpoint gate, 90 percentile): 2/3 fully clean (adamantyl_-Cl, adamantyl_-CF3); affinity-rank-1 1-indanyl_-Cl has 1 borderline flag (CYP3A4_Veith 90.4 vs gate 90.0). hERG band 65–69 percentile across class — no liability separation. → STRC h01 Phase 8c v5.3 ADMET-AI Triage 2026-04-26.

Lead committee (paper-grade tradeoff, no single winner):

Track	Lead	Strength	Weakness
Affinity-led	`1-indanyl_acylsulfonamide_SO2Me_-Cl`	Boltz-2 + Vina + n=20 mut-ensemble all rank-1	1 borderline ADMET flag (CYP3A4 90.4)
ADMET / kinetic-confidence	`adamantyl_acylsulfonamide_SO2Me_-Cl`	0 ADMET flags; tightest τRAMD SEM 0.65 ps	Mut-ensemble mean ΔG +4.09; relies on best-pose

Class-level verdict: v5.3 sulfonamide is the first head-class that passes the Vina Method-Class Rule — Anion Charge Magnitude vs Pocket Geometry (Vina × Boltz-2 AGREE) and the within-target lemma pre-screen (σ < 3×). It is the same as v5.2 on the cross-target kinetic-selectivity axis — both classes bounded ≤ ~3× ≪ 5× gate. The kinetic-selectivity rescue strategy at the head-group level is exhausted for adamantyl-/indanyl-cage scaffolds. v5.4 design rationale (next major version) must break the geometry coupling at the cage / linker, not at the head.

Paper claim sequence updated: §2.6 (off-target selectivity) is no longer “pending” — it is closed-by-bound for v5.2 and v5.3. The story arcs:

APBS proves the mechanistic asymmetry (§2.1)
Vina misses it (§2.2)
Pose-transplant + ensemble APBS recover it (§2.3, §2.4)
Coulomb-aware medchem yields adamantyl-cage v5 / v5.2 / v5.3 leads (§2.5)
Cross-target τRAMD demonstrates a within-target → cross-target spread lemma that bounds kinetic selectivity from a single-target measurement (§2.6, NEW) — saves ~5–10 h compute per scaffold class and provides a generalisable pre-screen for any cochlear off-target panel (TRPM4, KCNQ4, Cx50, BK, TMEM16A)
Lead-committee tradeoff (§2.7, NEW): no v5.3 ligand wins all axes simultaneously; affinity-track and ADMET/kinetic-track are reported separately

Outstanding before submission, updated:

5. Figures

WT vs mutant pocket isosurface (APBS φ) — placeholder
Figure 2 (rendered): phase5e_v2/paper_figure2.{png,svg} — Vina × Boltz-2 × τRAMD chemistry-class rule (anion ≥ −1.5 → AGREE / ≤ −2 → INVERT)
Figure 3 (rendered): phase5m_trpm4/paper_figure3.{png,svg} — STRC vs TRPM4 τRAMD cross-target kinetic-selectivity (v5.2 lead + anchor both FAIL gate)
Figure 4 (rendered): phase5q_strc_v53/paper_figure4.{png,svg} — v5.3 within-target τRAMD pre-emptive falsification via lemma (σ = 1.45× → bound 2.9× ≪ 5× gate)
v5 top combined leads, structural overlay, ADMET panel — placeholder
Lead-committee tradeoff matrix (NEW, paper-grade, render from Phase 8c v5.3 + Phase 5q): affinity-rank × τRAMD-rank × ADMET-flag-count grid for the v5.3 PASS-list — visualises the structural impossibility of “single perfect lead” in the v5.3 sulfonamide class

6. Outstanding before submission

Phase 5d-WT MD + Phase 5k-WT APBS (matched WT ensemble) → specificity claim
Phase 8d/8e v5.2 dock + top-30 (headcount expansion for n-power)
Phase 8f off-target selectivity panel on v5 top-5
Wet-lab validation (thermal shift, SPR Kd, patch-clamp hERG) — outsourced to collaborator lab
AF3 Phase 7H protein-ligand complex prediction for top-3 (Google auth blocker)
α sensitivity analysis (vary α in 0.5-2.0; confirm ranking stability)
Blinded retrospective test: does Coulomb-score predict known pharmacochaperones? (tafamidis/TTR, lumacaftor/CFTR)

Connections

[part-of] h01 hub
[complements] all Phase 4/5/8 proof notes above
[audience] Misha — primary motivation
[platform-for] future STRC Hypothesis Ranking extension to other STRC variants

STRC Research

Explorer

STRC Paper Draft Outline 2026-04-25