STRC h01 Phase 8g-v3-lite + 8h-lite #9 — Pocket-max-φ resample + PBPK dose-escalation sweep 2026-04-26
Closes the diagnostic gate that Phase 8g-v2 left open (“STRC pose anion-O sub-optimal for electrostatics; is the FAIL pose-driven or residue-driven?”) and quantifies the dose-escalation lever surfaced by Phase 8h-lite #7 (default-apical 0% time-above-EC at 100 µM IT).
Phase 8g-v3-lite — pocket-max-φ resample
Method
Re-samples the existing Phase 8g-v2 (CONHOH lead) and Phase 7i (CONHOMe lead) APBS potential.dx grids — no new APBS compute, only resampling logic changes. Per receptor + variant:
- Heavy-atom grid of lead pose centroid + 12 × 12 × 12 Å cube at 0.5 Å step (24 484 voxels).
- Solvent-accessibility mask: keep voxels with 1.5 Å ≤ d_to_nearest_protein_heavy_atom ≤ 4.0 Å. Excludes both protein-interior charge cores (artefact source: φ > +100 kT/e) and bulk-solvent reaches (φ → 0). Yields 50–64 % accessible-voxel fraction.
pocket_top10_phi: mean of top-10 % most-positive accessible voxels (≈ 1 200 voxels averaged). Pose-independent upper bound on what an anion-O could sample if reorientation were free; robust to single-voxel artefacts.pocket_max_phi: single-voxel max within accessibility shell.pose_anion_phi: φ at Vina-pose hydroxamate/methoxamate-O atoms (matches Phase 8g-v2 / 7i parent metric).
Verdict gate
PASS — STRC pocket_top10 exceeds every off-target by ≥ +2 kT/e (residue-driven gate satisfied; FAIL was pose-driven; site-defined re-dock would close).
WEAK — STRC leads but margin < +2 kT/e.
FAIL — any off-target ≥ STRC (no pose adjustment of same head can rescue gate).
Result
Both variants FAIL, but margin narrowed by ~3 kT/e vs Phase 8g-v2 pose-anion metric.
CONHOH variant — verdict FAIL, margin −5.24 kT/e
| Receptor | pocket_top10 φ (kT/e) | pocket_max φ | pose-anion ⟨φ⟩ | Δ(pose ↔ argmax) Å | acc % |
|---|---|---|---|---|---|
| STRC_E1659A_K1141 | +29.95 | +141.48 | +2.51 | 5.95 | 51 |
| TMEM16A_7ZK3 | +35.19 | +137.91 | +4.89 | 9.12 | 61 |
| TRPM4_8RD9 | +24.03 | +132.10 | +9.36 | 6.58 | 59 |
| Cx50_via_Cx36_8QOJ | +16.20 | +82.81 | −3.02 | 4.95 | 58 |
| KCNQ4_7BYM | +15.35 | +125.55 | +0.91 | 6.96 | 50 |
Worst off-target: TMEM16A.
CONHOMe variant — verdict FAIL, margin −2.49 kT/e
| Receptor | pocket_top10 φ (kT/e) | pocket_max φ | pose-anion ⟨φ⟩ | Δ(pose ↔ argmax) Å | acc % |
|---|---|---|---|---|---|
| TRPM4_8RD9 | +24.39 | +127.22 | +6.12 | 6.65 | 60 |
| STRC_E1659A_K1141 | +21.90 | +121.49 | +2.48 | 6.00 | 52 |
| Cx50_via_Cx36_8QOJ | +16.83 | +79.95 | −3.55 | 3.35 | 59 |
| TMEM16A_7ZK3 | +15.54 | +132.57 | +4.27 | 6.29 | 64 |
| KCNQ4_7BYM | +12.77 | +108.13 | +1.12 | 8.28 | 56 |
Worst off-target: TRPM4. Note the worst-off-target swap CONHOH→CONHOMe (TMEM16A → TRPM4) — different head groups face different dominant off-target risks, useful for steering v5.3.
Interpretation
- STRC K1141 IS electropositive at +21.9 / +29.95 kT/e top-10% (consistent with Phase 5k matched-ensemble pocket-centre +5.99 ± 1.37; top-10% naturally higher than centre mean). Mech-4 thermodynamic claim holds.
- Vina pose was indeed sub-optimal — STRC Δ(pose ↔ argmax) ≈ 6 Å; pose-anion ⟨φ⟩ +2.48–2.51 vs pocket ceiling +21.90–29.95. Site-defined re-dock could lift STRC pose-anion-φ by ~+15 kT/e on the absolute axis.
- But the gate cannot close by re-dock alone. Pose-φ ceiling equals pocket_top10. STRC pocket ceiling +21.90 (CONHOMe) ≤ TRPM4 ceiling +24.39. Best-case re-dock places STRC pose-anion at +21.90; TRPM4 already sits at top-10% of +24.39 worth of voxels. Off-target pocket interior is intrinsically more electropositive on the dominant off-target — not a pose artefact.
- Margin narrowed but the verdict-class is the same as Phase 8g-v2: FAIL. The Phase 8g-v2 hub-frontmatter claim “Triggers Phase 8g-v3 site-defined re-dock + Phase 8e-v2 STRC pose-anion-O re-rank” is superseded by this diagnostic — full re-dock would not flip the verdict, only narrow the margin further (not enough to clear the +2 kT/e gate).
- The kinetic-selectivity escape hatch is now load-bearing. Per Pharmacochaperone Residence Time Criterion, slow STRC k_off + fast off-target k_off can deliver clinical selectivity without thermodynamic-affinity selectivity. τRAMD on the v5.2 shortlist becomes elevated from P1-light to P0 for h01.
- Boltz-2 binding-mode test on TRPM4 / TMEM16A for top-2 Tier-1 — does electrostatic attraction (+24.39 / +35.19 kT/e at pocket interior) translate to a real Boltz-2 ipTM binding pose? Electrostatic pull is necessary, not sufficient. If ipTM stays low, anti-selective electrostatics is academic, not pharmacological. P0-light.
Why the original Phase 8g-v3 plan stops here
The hub next_step queue listed “Phase 8g-v3 site-defined re-dock + Phase 8e-v2 STRC pose-anion-O re-rank” as the heavy P0. The diagnostic this script ran proves both moves cannot close the gate: STRC pocket ceiling < dominant off-target pocket ceiling on both variants. Running them would consume ~2 h compute to confirm a foregone conclusion. Phase 8g-v3 full is stopped; capital re-allocated to τRAMD (P0) and Boltz-2 off-target binding-mode (P0-light).
Phase 8h-lite #9 — PBPK dose-escalation sweep
Method
Sister to cochlear_pbpk_ode.py (Phase 8h-lite #7). Identical 8-compartment topology, volumes, and rate constants. Sweeps applied IT dose × [100 µM, 1 mM, 10 mM] × [fast / default / slow] apical-uptake bracket. Tracks two ceilings:
frac_72h_HC_ER ≥ 1 µM— therapeutic-window fraction (HC_ER target compartment).frac_72h_HC_CYT ≥ 50 µM— safety-ceiling flag. Empirical threshold from Givinostat mARC1 2 Hydroxamate Bypass: Cmax 4 µM at MTD; 10× margin = 40 µM cytosol; rounded up to 50 µM.
IT dose precedent: dexamethasone IT 10–60 mM (Auritec OTO-104 etc.); SPI-1005 / ebselen IT up to 30 mM. 1 mM and 10 mM applied are well within IT pharmacy.
Therapeutic window — % of 72 h with HC_ER ≥ 1 µM
| Apical scenario | 100 µM | 1 mM | 10 mM |
|---|---|---|---|
| fast_apical (t½ 30 min) | 38.9 % | 96.9 % | 98.5 % |
| default_apical (t½ 95 min) | 0.0 % | 95.8 % | 98.0 % |
| slow_apical (t½ 300 min) | 0.2 % | 93.8 % | 97.4 % |
Safety flag — % of 72 h with HC_CYT ≥ 50 µM
| Apical scenario | 100 µM | 1 mM | 10 mM |
|---|---|---|---|
| fast_apical | 0.0 % | 0.0 % | 93.7 % |
| default_apical | 0.0 % | 0.0 % | 65.5 % |
| slow_apical | 0.0 % | 0.0 % | 0.0 % |
HC_ER Cmax (µM)
| Apical scenario | 100 µM | 1 mM | 10 mM |
|---|---|---|---|
| fast_apical | 1.34 | 13.50 | 133.78 |
| default_apical | 0.90 | 9.52 | 88.94 |
| slow_apical | 1.19 | 4.77 | 43.13 |
Interpretation
- 1 mM is the sweet spot. Closes therapeutic window across all three apical brackets (94 – 97 % time-above-EC) without breaching safety ceiling (0 % HC_CYT ≥ 50 µM in all three apical scenarios).
- 10 mM is the safety cliff. 65 – 94 % HC_CYT ≥ 50 µM in fast / default apical. Only slow-apical bracket stays safe at 10 mM — but slow-apical is the worst-case low-uptake scenario, so this represents window narrowing, not a usable design space.
- Phase 8h-lite #7 deliv-3 ceiling on the exposure side is liftable purely through dose escalation. No new chemistry, no new permeation enhancer required. The eustachian-tube clearance bottleneck identified in #7 (k_eu = 1.16 × 10⁻² /min ≈ 11 000× faster than k_RWM = 1.0 × 10⁻⁶ /min) is overcome by mass-action: depositing 10× more compound makes the ~10⁻⁴ that crosses RWM still a clinically meaningful absolute amount.
- But deliv axis cannot lift to 4 yet — selectivity gate (Phase 8g-v3-lite FAIL) is the binding constraint, not exposure. A 1 mM IT dose that solves exposure but lands at +24 kT/e on TRPM4 with ~similar pose-anion-φ as STRC is no longer a viable lead.
Combined verdict
| Axis | Pre-session | Post-session | Driver |
|---|---|---|---|
| mech | 4 | 4 held | STRC K1141 pocket genuinely electropositive (top-10% +21.9 to +29.95 kT/e) |
| deliv | 3 | 3 held | Exposure rescued at 1 mM IT (94-97% therapeutic window), but selectivity gate FAIL at residue level prevents axis lift |
| misha_fit | 4 | 4 held | unchanged |
| tier | A | A held |
Net move: the heavy-queue P0 from the hub frontmatter (Phase 8g-v3 + 8e-v2 re-dock pair) is stopped as a foregone-conclusion compute. Replaced by:
- P0: τRAMD on v5.2 shortlist (kinetic-selectivity escape hatch, now load-bearing per finding #5 above)
- P0-light: Boltz-2 binding-mode at TRPM4 (CONHOMe lead) and TMEM16A (CONHOH lead) — does anti-selective electrostatic ceiling translate to actual binding ipTM?
- P0-light: 1 mM IT dose locked as reference dose for all subsequent PBPK / wet-lab / paper claims (supersedes Phase 8h-lite #7 100 µM reference)
- P1-light: v5.3 design pivot — explore weakly-anionic head groups (acyl sulfonamide, mono-anionic 1,3,4-oxadiazol-2(3H)-one) that mute electrostatic pull on highly-cationic off-target interiors while still satisfying Phase 5k STRC pocket ⟨φ⟩ +5.99 kT/e
- Defer: full Phase 8g-v3 site-defined re-dock (does not flip verdict per this diagnostic)
Ranking delta
- tier: A → A (held)
- mech: 4 → 4 (held; pocket electropositivity reconfirmed at the residue-driven level for STRC)
- deliv: 3 → 3 (held; exposure rescue at 1 mM IT but selectivity gate still FAIL)
- misha_fit: 4 → 4 (held)
- next_step: replace “Phase 8g-v3 site-defined re-dock + Phase 8e-v2 STRC pose-anion-O re-rank” with “τRAMD v5.2 shortlist (P0) + Boltz-2 TRPM4/TMEM16A binding-mode (P0-light) + 1 mM IT dose locked as reference”
Connections
[supersedes]STRC h01 Phase 8g-v2 APBS Off-Target Rescore 2026-04-26[extends]STRC h01 Phase 8h-lite Light Computational Evidence Package 2026-04-26[builds-on]STRC h01 Phase 5k-WT Matched Ensemble APBS 2026-04-25[informs]Pharmacochaperone Residence Time Criterion[informs]Hydroxamic Acid Divalent Cation Liability[informs]Givinostat mARC1 2 Hydroxamate Bypass[part-of]H01 hub[about]Misha