STRC h01 Phase 6c hERG Extension
21 of 28 ligands hit hERG at sub-10 µM Kd. Every v3b/v4 top-10 tight binder docks to hERG with Kd 0.3-2.4 µM — 3-14× tighter than to K1141. All 3 ADMET-clean Phase 4h Tier-1 candidates also hit hERG: benzox-3OH-4F-COOH 8.8 µM, indole-3OMe 3.2 µM, benzimid-3OMe-COOH 7.1 µM. Only tafamidis, benzoxazole-3OMe-COOH, benzoxazole-3CN-COOH, diflunisal, meclofenamic, aspirin, DMSO clear the 10 µM threshold. This is a class liability of the bicyclic-aromatic-COOH pharmacophore at the astemizole-analogous hERG pore-blocker site — affinity optimisation at K1141 automatically increases hERG affinity. Systemic oral route is not viable for this chemistry; intracochlear delivery (RWM / intratympanic) remains open. FDA-mandatory gate G2.5 in Phase 8 SOP should now be hERG + TMEM16A + COX.
Problem
STRC h01 Phase 6c Full 6-Target Panel 2026-04-24 (morning 2026-04-24) added TRPM4, Cx50, KCNQ4, TMEM16A, COX-1, COX-2 to the cochlear off-target panel and found a TMEM16A class liability and a COX class liability. hERG was explicitly deferred — no NSAID/blocker-bound hERG PDB was curated in the 6-target batch.
hERG (KCNH2) is the FDA-mandatory cardiac-safety gate. No drug development track can proceed without clearing it. Even for the cochlear-local delivery route we need to know the residual systemic leak hERG exposure.
Method
- Tool: Phase 6c pipeline (
pharmacochaperone_phase6c_offtarget_panel.py) extended with a 7th target; driver scriptpharmacochaperone_phase6c_herg_extension.pyfor the isolated hERG run without re-docking the existing 6 targets. - Target: PDB 8ZYO — human hERG (KCNH2) cryo-EM 3.29 Å, astemizole-bound (HETATM XB7), Miyashita 2024 Structure 32:1926 (DOI 10.1016/j.str.2024.08.021). CCD verified via RCSB text_chem API 2026-04-24.
- Box: 22 × 22 × 22 Å, centred on astemizole centroid (156.16, 151.30, 132.27). Larger than h01 K1141 box because channel pockets vary.
- Vina v1.2.7, exhaustiveness 16, num_modes 5, cpu 4 — identical to existing Phase 6c settings.
- Ligand set (28 unique SMILES after dedup from 33 candidates):
- 10 DEFAULT_PROBES (tafamidis + 4 fenamic parents + 2 benzoxazole-COOH + 2 negatives + DMSO)
- 10 v3b top-10 (ensemble-mean ΔG)
- 10 v4 top-10 (ensemble-mean ΔG)
- 3 ADMET-clean Phase 4h Tier-1 (benzox-3OH-4F-COOH, indole-3OMe, benzimid-3OMe-COOH)
Results
Aggregate
| hERG Kd band | n | interpretation |
|---|---|---|
| < 10 µM (hit) | 21/28 | FDA-mandatory cardiac risk |
| 10-100 µM (borderline) | 5/28 | manageable with monitoring |
| ≥ 100 µM (clean) | 2/28 | aspirin + DMSO controls |
Top hERG hitters (tightest — worst)
| rank | ligand | hERG Kd | h01 Kd | selectivity h01:hERG | class |
|---|---|---|---|---|---|
| 1 | nc__3-amino-benzofuran-2-COOH__1-naphthyl__CONHOH__-CF3 | 0.31 µM | 4.6 µM | 0.07 | v3b top #1 |
| 2 | v4__3-amino-benzofuran-2-COOH__6-F-2-naphthyl__CONHOH__-F | 0.50 µM | 5.8 µM | 0.09 | v4 top #6 |
| 3 | nc__3-amino-benzoic__4-F-biphenyl__tetrazole__-CF3 | 0.51 µM | 4.9 µM | 0.10 | v3b top #3 |
| 4 | v4__3-amino-benzofuran-2-COOH__6-F-2-naphthyl__tetrazole__-F | 0.57 µM | 4.8 µM | 0.12 | v4 top #2 |
| 5 | nc__3-amino-benzofuran-2-COOH__4-F-biphenyl__tetrazole__-Me | 0.67 µM | 5.0 µM | 0.13 | v3b top #4 |
All v3b/v4 top-10 tight binders show hERG Kd < 3 µM. Selectivity ratios 0.07-0.41 — every single one preferrs hERG over K1141 by 3-14×.
Cleanest ligands
| ligand | hERG Kd | interpretation |
|---|---|---|
| DMSO | 35,314 µM | solvent control — passes |
| aspirin | 156 µM | negative control — passes |
| meclofenamic | 27.1 µM | borderline |
| diflunisal | 17.1 µM | borderline (matches literature mild hERG activity) |
| benzoxazole-3CN-COOH | 14.6 µM | borderline |
| tafamidis | 12.4 µM | borderline (below FDA-clearance margin) |
| benzoxazole-3OMe-COOH | 12.3 µM | borderline |
Phase 4h ADMET-clean Tier-1 candidates
| candidate | hERG Kd | verdict |
|---|---|---|
| benzox-3OH-4F-COOH (#15) | 8.8 µM | FAIL sub-10 µM |
| indole-3OMe (#17) | 3.2 µM | FAIL sub-10 µM |
| benzimid-3OMe-COOH (#27) | 7.1 µM | FAIL sub-10 µM |
The 3 compounds that cleared Phase 6c cochlear off-target + scaffold tox audit + ADMET-AI now fail hERG at sub-10 µM Kd. There are zero compounds across the full h01 shortlist that simultaneously clear all four gates.
Pattern interpretation — class liability, not compound liability
The hERG hit rate is 21/28 = 75 %. For context, the earlier TMEM16A panel hit rate was 10/10 on v4 top compounds and COX hit rate was 10/10. The common structural feature is the fused-aromatic + polar-group + lipophilic-tail scaffold, which happens to match the astemizole-class hERG pore-blocker pharmacophore:
- Aromatic core: fits the aromatic cage of the hERG pore cavity (F656 / Y652 in KCNH2)
- Polar head (COOH / CONHOH / tetrazole): fits the cationic-gate / T623 interaction
- Lipophilic tail (CF3 / naphthyl / biphenyl): fits the hydrophobic extension pocket
This is not a compound-level problem — it is a scaffold-level problem. Every time we push the chemistry toward higher K1141 affinity, we simultaneously increase hERG affinity because both pockets recognise the same pharmacophore.
The only structurally divergent chemistry in our shortlist is:
- tafamidis (rigid bicyclic + tilted dichlorophenyl — 12.4 µM, borderline)
- benzoxazole-3OMe/3CN-COOH (lacks extended tail — 12.3-14.6 µM, borderline)
None of these are in our tight-binding tier (all > 30 µM at K1141).
Ranking delta
- Hypothesis h01: tier A held | mech 2 held | deliv 4 → 3 (systemic oral route essentially closed for this chemistry class; intracochlear / RWM delivery required) | misha_fit 4 held
- Next-step update:
- Delivery axis: cochlear-local delivery (intratympanic / RWM) moves from option to requirement. Phase 4 plan compartmental PK on RWM→endolymph→OHC must confirm that peak systemic concentration stays below hERG IC50 at the highest cochlear-effective dose.
- Chemistry axis: v5 library design must explicitly include a hERG-selectivity filter. Candidate strategies:
- Reduce lipophilic tail (remove -CF3, naphthyl, biphenyl); constrain MW ≤ 300
- Introduce basic nitrogen (amine with pKa 7-9) — known to disrupt hERG binding
- Explore non-planar rigid scaffolds (tafamidis-class) at slight affinity cost
- Phase 8 SOP update (if/when Phase 8 is back on deck): G2.5 kill-gate now COX + TMEM16A + hERG (adds hERG to the existing two).
Connections
[part-of]h01 hub[see-also]STRC h01 Phase 6c Full 6-Target Panel 2026-04-24[see-also]STRC h01 Phase 6c COX Selectivity Panel 2026-04-24[see-also]STRC h01 Phase 7 ADMET-AI Triage 2026-04-24[contradicts]STRC h01 Phase 4h Tafamidis Playbook Library 2026-04-23 — the bioisosteric optimisation direction intended to kill COX also fails to kill hERG; rationale needs revision.[source]Miyashita 2024 Structure 32:1926 — primary paper for 8ZYO (paper note not yet ingested; TODO)