STRC h01 Phase 3c v2 Expanded Screen 2026-04-23
TL;DR. 667-ligand expanded screen (619 combinatorial scaffold × acid bioisostere + 29 curated FDA carboxylates) against K1141 with ensemble docking across 5 MD-sampled receptor conformers. Ceiling reached: niflumic / flufenamic / sulfasalazine at Kd ≈ 30 μM, f_PC ≈ 0.125 — 1.7× better than Phase 5b best (diflunisal 50 μM, f_PC 0.083) but still 8× short of NORMAL-rescue threshold f_PC ≥ 0.50. Verdict: RED overall, but fenamic-acid scaffold family revealed as a new direction (2-(arylamino)benzoic acids beat Phase 4b leads by +0.35 kcal/mol). Tetrazole bioisostere of COOH competitive. Next: Phase 3c v3 ZINC22 bioactives library (~20 k) + Phase 3c v4 fragment-growing on niflumic/sulfasalazine cores, OR pivot to covalent/reversible-covalent chaperone strategy targeting K1141 Lys-NH₃⁺.
Rationale
Phase 5b RED-LIGHT (all Phase 4b leads + diflunisal f_PC < 0.10) + Phase 5c K1141 site GREEN-LIGHT (pocket structurally stable, no better cryptic site) converged on: chemistry is the bottleneck. Phase 3c v2 expands the chemical space in three directions: (a) acid bioisosterism, (b) aromatic/heteroaromatic scaffold diversity, (c) FDA-approved carboxylate drugs as high-value anchors.
Method
Implemented as two scripts:
-
pharmacochaperone_phase3c_v2a_library_build.py— RDKit combinatorial enumeration:- 25 scaffolds: benzene, toluene, phenol, anisole, aniline, halobenzenes, naphthalene, biphenyl, pyridine/pyrimidine/pyrazine, thiophene, furan, oxazole, imidazole, quinoline, isoquinoline, indole, benzimidazole, benzoxazole, benzothiazole, benzofuran, benzothiophene, indoline-2-one
- 10 acid / acid-bioisostere groups: -COOH, -CH₂COOH, -OCH₂COOH, -CONHOH (hydroxamic), -SO₂NH₂ (sulfonamide), -CONHSO₂CH₃ (acylsulfonamide), -PO(OH)₂ (phosphonate), 1H-tetrazole, -CH=CHCOOH (acrylic), -NHC(O)COOH (oxamic)
- Attached at each aromatic H-bearing carbon; canonical-SMILES dedup → 828 raw → Lipinski + has-acid filter → 619 drug-like combinatorial + 29 curated drugs (Phase 4b leads + fenamic NSAIDs + TTR analogs + diuretics + chaperone precedents) = 667 ligands in 3D PDBQT.
-
pharmacochaperone_phase3c_v2b_ensemble_dock.py— two-stage docking:- Stage 1 breadth: Vina exh 8 × 3 modes against snap_008 (global-pocket rep) on all 667 ligands → ~1.5 lig/s → 8.5 min, top 30 by ΔG.
- Stage 2 ensemble: Vina exh 16 × 5 modes on top-30 × 5 k-means-selected receptor conformers (snap_006, 012, 015, 017 + global rep 008; k-means on 9-residue K1141 pocket Cα flat-vector) → mean ± std ΔG → Kd(μM) = exp(ΔG/RT) × 10⁶ → bound fraction θ = [L]/([L]+Kd) at [L] = 10 μM → f_PC = θ × 0.5 (RESCUE_ETA 0.5, conservative per Phase 5b convention).
Box: Phase 4b K1141 centre (-22.027, -18.547, 2.215), 18×18×18 Å.
Results
Stage 1 top-10 (single conformer snap_008, exhaustiveness 8)
| # | Name | ΔG (kcal/mol) |
|---|---|---|
| 1 | sulfasalazine | -6.80 |
| 2 | tafamidis-analog (quinoline-3-carboxylic 6-phenoxy) | -6.43 |
| 3 | niflumic-acid | -6.40 |
| 4 | naphthalene-tetrazole | -6.36 |
| 5 | flufenamic-acid | -6.34 |
| 6 | biphenyl-CONHSO₂Me | -6.31 |
| 7 | quinoline-NHC(O)COOH | -6.25 |
| 8 | indoline-2-one-tetrazole | -6.22 |
| 9-10 | biphenyl-tetrazole (2 positions) | -6.22, -6.21 |
Stage 2 ensemble rescoring (top-30 × 5 conformers, exh 16)
| # | Name | Mean ΔG ± std (kcal/mol) | Kd (μM) | f_PC at [L]=10 μM |
|---|---|---|---|---|
| 1 | niflumic-acid | -6.18 ± 0.46 | 29.5 | 0.127 |
| 2 | flufenamic-acid | -6.17 ± 0.41 | 30.0 | 0.125 |
| 3 | sulfasalazine | -6.15 ± 0.49 | 30.7 | 0.123 |
| 4 | biphenyl-tetrazole-1 | -6.06 ± 0.37 | 36.0 | 0.109 |
| 5 | naphthalene-tetrazole-0 | -6.04 ± 0.27 | 37.3 | 0.106 |
| 6 | tafamidis-analog | -5.97 ± 0.49 | 41.8 | 0.097 |
| 7 | ketoprofen | -5.93 ± 0.32 | 44.7 | 0.091 |
| 8 | naphthalene-CONHSO₂Me-0 | -5.92 ± 0.30 | 45.3 | 0.090 |
| 9 | isoquinoline-CONHOH-4 | -5.91 ± 0.24 | 46.2 | 0.089 |
| 10 | naphthalene-tetrazole-2 | -5.88 ± 0.25 | 48.8 | 0.085 |
Verdict: 0 GREEN (f_PC ≥ 0.50), 0 YELLOW (0.25 ≤ f_PC < 0.50), 30 RED.
Delta from Phase 5b baseline
| Metric | Phase 5b best | Phase 3c v2 best | Improvement |
|---|---|---|---|
| Best lead ΔG (ensemble) | -5.87 kcal/mol (diflunisal pos. control) | -6.18 kcal/mol (niflumic-acid) | +0.31 kcal/mol |
| Best Kd | 50 μM | 29.5 μM | 1.7× tighter |
| Best f_PC at [L]=10 μM | 0.083 | 0.127 | 1.5× higher |
| Chemical scaffolds | 5 aryl-COOH | 2-aryl-aminobenzoic + azo + bicyclic | qualitative expansion |
Chemistry insights
- Fenamic acid family wins: niflumic (2-(3-CF₃-phenyl-amino)nicotinic), flufenamic (2-(3-CF₃-phenylamino)benzoic), and meclofenamic/tolfenamic/mefenamic cluster near top. These are NSAIDs with anthranilic acid backbone + arylamine extension — adds aromatic stacking + H-bond donor/acceptor versatility vs Phase 4b’s simpler aryl-COOH. Direction: further extend the arylamine side.
- Tetrazole bioisostere competitive with COOH: biphenyl-tetrazole at -6.06 beats biphenyl-COOH equivalents in our library. Tetrazole’s flat 5-ring + 4 HBA nitrogens fits pocket electrostatics as well as carboxylate. Useful for CNS/oral PK if needed.
- Scaffold size matters: biphenyl > naphthalene ≥ indole > benzene by 0.2-0.5 kcal/mol on average. K1141 pocket accommodates MW ~250-350 better than MW ~150-200.
- Acylsulfonamide underwhelms: biphenyl-CONHSO₂Me mid-pack (#4 stage 1 but #6 stage 2), lost ground in ensemble rescoring. Suggests the pose is conformation-sensitive.
- Hydroxamic, phosphonate, oxamic: no breakout hits. Charge/size mismatch at pocket K1141.
The ceiling problem
Phase 3c v2 best ensemble mean ΔG is -6.18 kcal/mol (niflumic-acid). To reach f_PC ≥ 0.50 at [L]=10 μM, we need:
Required Kd = [L] × (f_PC / η) / (1 - f_PC/η)
= 10 × 1.0 / 0 (asymptote)
(at f_PC = 0.50, η = 0.5 → θ must equal 1.0, infeasible)
Re-examining: with η = 0.5, max f_PC = 0.50 × θ_max = 0.50. So f_PC = 0.50 requires θ = 1.0 (100% bound), which requires Kd → 0.
The conservative assumption η = 0.5 (half of bound molecules give functional rescue) caps f_PC at 0.50 no matter how tight the Kd. To cross f_PC = 0.50 we need either (a) η > 0.8 (which requires wet-lab data on PC rescue efficiency per bound molecule), or (b) [L] > 10 μM therapeutic (may trigger off-target).
More honest target: f_PC = 0.30 (MILD-MODERATE rescue in Misha stack model). At η = 0.5, this requires θ = 0.60 → Kd = [L]×0.40/0.60 = 6.7 μM. ΔG ≈ -7.06 kcal/mol — 0.88 kcal/mol below current best.
Routes to -0.88 kcal/mol improvement:
- Larger library + deeper sampling (Phase 3c v3 ZINC22 bioactives ~20k): expected +0.5-1.0 kcal/mol
- Fragment growing on niflumic core (Phase 3c v4): expected +1.0-1.5 kcal/mol if K1141 salt bridge established
- Reversible covalent warhead at Lys-NH₃⁺ (Phase 6b covalent strategy): +2-3 kcal/mol but selectivity risk
- FEP/TI rescoring (Phase 6a): not a Kd improvement, just a noise reduction on the top hits
Reframe for Misha stack model
Phase 3c v2 suggests h01 monotherapy can at best deliver MILD-MODERATE rescue (f_PC 0.25-0.35) at Kd ~5-10 μM and η ~0.5. Reaching NORMAL (f_PC ≥ 0.50) requires either:
- A substantially tighter Kd (< 1 μM) from Phase 3c v3 expanded screen — achievable in principle but low probability at current chem-space growth rate, OR
- A higher η (> 0.8) — unknown until wet-lab rescue assay, structurally dependent, not computationally estimable yet.
Updated clinical framing: the Misha Compound-Het Therapy Stack Model projection that “h01 is the only monotherapy route to NORMAL” is compute-uncertain. h03 AAV remains the S-tier primary lever. h01 now frames as:
- MILD-MODERATE adjunct lever (highly viable): adds 10-20 dB ABR improvement on top of h03’s 20-30 dB, stacking to near-NORMAL.
- Potential NORMAL monotherapy (uncertain): conditional on Phase 3c v3/v4 finding sub-μM lead AND wet-lab η > 0.8.
Ranking delta
- #1 (h01) Pharmacochaperone E1659A: A held. Mech 3 held, Deliv 4 held, Misha-fit 4 held. Phase 3c v2 delivered partial progress (1.7× Kd improvement from Phase 5b) but no GREEN. Next-step reordered:
- Phase 3c v3 ZINC22 bioactives + DrugBank FDA full (~20 k ligands) — one-day Mac compute.
- Phase 3c v4 fragment-growing on niflumic/fenamic core — K1141 salt-bridge engineering.
- Phase 6b reversible covalent strategy (parallel) — K1141 Lys-NH₃⁺ warhead (acrylamide, β-keto amide).
- #28 Misha Compound-Het Therapy Stack Model: ref-entry updated — h01 reframed from “NORMAL monotherapy pathway” to “MILD-MODERATE adjunct with conditional NORMAL path”. h03 remains S-tier primary lever.
What this doesn’t prove
- Library ceiling vs Kd ceiling: our 667-library is small. ZINC22 bioactives (~180 k) contains 250× more chemical diversity. A sub-μM lead could well exist — not found here.
- Vina bias on carboxylates: Vina empirical scoring under-rewards negatively-charged ligands interacting with lysines (no explicit electrostatic term). Phase 5b noted this. Fenamic acids docking to Lys-rich pocket may be 0.5-1.0 kcal/mol under-scored.
- η = 0.5 assumption: conservative placeholder. Real η depends on whether PC stabilizes the productive folded state vs any bound state; only wet-lab thermal shift / aggregation rescue can calibrate.
- 2 ns MD may undersample pocket flexibility: 20 ns extension cheap on Mac; would help confirm the 0.13 ceiling is MD-converged not trajectory-limited.
Connections
- STRC h01 Phase 5 MD Ensemble Rescoring 2026-04-23 — Phase 5b RED-LIGHT
- STRC h01 Phase 5c Cryptic Pocket Analysis 2026-04-23 — K1141 GREEN-LIGHT
- STRC Pharmacochaperone Virtual Screen E1659A — main hypothesis
- Misha Compound-Het Therapy Stack Model — reframed h01 role
[applies]STRC Pharmacochaperone K1141 Fragment Pocket — target site[part-of]STRC Hypothesis Ranking Log[about]Misha