STRC h01 Phase 6c COX Selectivity Panel 2026-04-24

Off-target selectivity dock against COX-1 (1Q4G, bound bromoflufenamic / BFL) and COX-2 (5IKR, bound fenamic-class ligand / ID8) on the Phase 4h Tier-1 commercial set + parent fenamate probes. Box centroids derived from bound-ligand HETATM coordinates (no fabricated pocket coordinates).

Pipeline: download → strip waters/ions → PDBQT → vina exh=16 num_modes=5 cpu=4 (left 4 CPUs for v4 production). 10 ligands × 2 targets = 20 docks delivered in 1 min 36 s.

Receptor prep status — Stage A

TargetPDBStatusBox source
COX-25IKR✅ dockablebound ID8 ligand centroid (38.96, 2.35, 61.50)
COX-11Q4G✅ dockablebound BFL ligand centroid (26.62, 33.70, 201.93)
TRPM47MF0❌ fetch_failed (404)wrong PDB ID — needs replacement
Cx507JN0⚠ prepped, no boxapo cryoEM; needs lit-cited fenamate-binding residues
BK channel6V3G⚠ prepped, no boxapo or PIP2-only; needs Tao 2017 site lookup
KCNQ47BYL⚠ prepped, no boxneeds Li 2021 retigabine-pocket residues
TMEM16A5OYB⚠ prepped, no boxneeds Paulino 2017 NPPB-pocket residues

Channel panel is deferred — each requires reading the cited primary paper and curating chain-A pocket residue numbers. Per literature-first discipline, no fabricated boxes added.

Stage B — COX-1 / COX-2 dock results

COX-2 (5IKR, fenamic-class bound, fenamic Arg120/Tyr385/Ser530 site)

Ligandmean ΔGKd (µM)Phase 5e h01 mutant KdSelectivity h01 / COX-2
niflumic-acid (parent)−8.940.2841.10.007 ← prefers COX 147×
flufenamic-acid (parent)−8.770.3732.70.011 ← prefers COX 88×
benzoxazole-3OMe-COOH (4h #2)−8.190.99(not in 5e)
diflunisal (5e positive)−8.141.08104.20.010 ← prefers COX 96×
tafamidis (4h #5)−7.951.47(not in 5e)
benzoxazole-3CN-COOH (4h #1)−7.851.76(not in 5e)
iododiflunisal−7.542.97(not in 5e)
meclofenamic-acid (parent)−7.503.1894.00.034
aspirin (negative ctrl)−6.3223.2
DMSO (proxy negative)−2.5314,035

COX-1 (1Q4G, bromoflufenamic-bound)

Ligandmean ΔGKd (µM)
niflumic-acid−8.700.42
flufenamic-acid−8.580.51
diflunisal−8.340.77
meclofenamic-acid−8.300.82
benzoxazole-3OMe-COOH−8.200.97
benzoxazole-3CN-COOH−8.171.02
tafamidis−7.921.54
iododiflunisal−6.918.49
aspirin−6.0040.0
DMSO−2.4216,929

Pipeline calibration ✅

Parent fenamate predicted Kds match published COX IC₅₀s within an order of magnitude:

  • niflumic acid published COX-2 IC₅₀ ≈ 0.1–0.3 µM, predicted 0.28 µM ✅
  • flufenamic acid published COX-2 IC₅₀ ≈ 0.5–3 µM, predicted 0.37 µM ✅
  • aspirin negative-control comes in at 23 µM (real aspirin acetylates COX irreversibly so equilibrium Vina underestimates; expected since Vina has no covalent modeling)
  • DMSO floor at 14 mM ✅ (no specific binding)

Pipeline is honest. Predicted Kds are interpretable.

Real finding — tafamidis-playbook bicyclic-core thesis is contradicted

Phase 4h table claim (Core A, “tafamidis #5”): “Tafamidis no COX liability.” Design thesis: “Bicyclic fusion breaks the COX Arg120/Tyr385 geometry.”

Phase 6c result:

  • Tafamidis itself docks at COX-2 Kd 1.5 µM and COX-1 Kd 1.5 µM
  • Benzoxazole-3CN-COOH (Phase 4h Tier-2 lead, “kill TRPM4 + COX”) docks at COX-2 Kd 1.8 µM and COX-1 Kd 1.0 µM
  • Benzoxazole-3OMe-COOH docks at COX-2 Kd 0.99 µM and COX-1 Kd 0.97 µM

The bicyclic benzoxazole-COOH pharmacophore is NOT a COX-killing modification as docked. The Arg120 salt bridge to the carboxylate is preserved; the bicyclic body fits the COX hydrophobic channel just as well as the anthranilic NHAr body.

Selectivity ratios for fenamic parents tell the same story: niflumic / flufenamic / diflunisal all prefer COX-2 over h01-K1141 by 88–147× — that’s why parent fenamates are NSAIDs, not pharmacochaperones. The Phase 5e flufenamic h01 Kd 32.7 µM exists in the same docking pipeline as Phase 6c flufenamic COX-2 Kd 0.37 µM, on calibrated boxes. Translation: parent fenamates and bicyclic-COOH bioisosteres ALL prefer COX over h01 K1141 by ≥30×, even on the v3b YELLOW chemistry.

Real-world implication for Misha’s pediatric target: any compound carrying the COOH + bicyclic aromatic combination requires explicit COX-deselection design beyond what Phase 4h Tier-1/Tier-2 proposed. The “bicyclic = no COX” thesis is wrong as docked.

What this means for the path forward

  1. Phase 4h playbook needs revision (separate edit). The “Tier-1 commercial → order this week” plan is still valid for calibration of ThermoFluor/MST/cell rescue assays, but it should NOT be presented as a low-tox orderable lead set. Tafamidis & iododiflunisal will register on a COX activity assay.
  2. Phase 8 SOP COX risk needs upgrading — the parallel cochlear ion-channel panel must include COX-1/-2 activity (CellSensor or PGE2 ELISA) as a G2.5 kill gate, not deferred to G3 tox.
  3. v4 fragment-grow (running) needs a COX-deselection sub-step. Top-N v4 hits with retained -COOH and bicyclic core will dock at COX. Real medchem differentiation requires either:
    • Replace COOH with non-acidic chaperone-pharmacophore (e.g., neutral H-bond acceptor like sulfone) — risky for K1141 salt bridge, needs testing
    • Add steric clash group at the position blocked by COX-Tyr385 hydroxyl — geometry analysis of best v4 hit aligned to ID8 in 5IKR could reveal a position
    • Remove the proximal aromatic ring (e.g., aliphatic linker) — kills planarity, kills COX, but kills π-stacking pocket fit too
  4. Channel panel still required. TRPM4 / Cx50 / BK / KCNQ4 / TMEM16A boxes need primary-paper lit curation in a follow-up turn (each is 30 min of literature reading + JSON edit). Until then, Phase 8 G3 cochlear-channel patch-clamp is the only quantitative kill gate for those targets.

Action items — not done in this turn

  • Replace TRPM4 7MF0 with a valid PDB (likely 6BCO or 6BCL — check)
  • Curate Cx50 / BK / KCNQ4 / TMEM16A box centres from primary literature (Flores 2020 / Tao 2017 / Li 2021 / Paulino 2017)
  • Re-run —dock with the curated 5-target additions
  • Phase 4h playbook note — add “COX selectivity contradicted” caveat and revise Tier-1 framing
  • Add 1 paragraph to Phase 8 SOP G2.5 COX activity gate

Ranking delta

Tier A held. mech 3 / deliv 4 / misha_fit 4 all held.

  • Phase 4h “bicyclic core kills COX” design thesis falsified for tafamidis + 3CN/3OMe benzoxazole bioisosteres. This is a material design constraint, not a tier change — pharmacochaperone monotherapy/adjunct framing per Misha Compound-Het Therapy Stack Model still holds, but the chemistry pipeline now needs explicit COX-deselection beyond what Phase 4h proposed.
  • The Phase 5e WT-pocket-validated docking pipeline is now also COX-calibrated (parent fenamates within 1 order of magnitude of published IC₅₀s), giving us a self-consistent selectivity scoring system across h01 and at least 2 off-targets.
  • next_step unchanged: Phase 3c v4 still primary; v4 verdict will inform whether a COX-deselect sub-task gets a Phase 3c v4b (fragment-replace COOH cluster) or whether enough v4 leads have alternative pharmacophores to skip it.
  • lit_audit: fixed held; this is computed on prepared receptors, no new literature claims.

Connections