STRC h01 Phase 3c v4 Fragment-Grow Delivery
Phase 3c v4 fragment-grow on the v3b 3-amino-benzofuran-2-COOH winner cluster + Phase 4h tafamidis-playbook seeds delivered 0 GREEN / 32 YELLOW / 18 RED over a 1,572-ligand library. Top mean ΔG = −7.27 kcal/mol (Kd 4.65 µM, f_PC 0.341) — within 0.01 kcal/mol of the v3b ceiling (−7.28, 4.57 µM, 0.343). Fragment-grow did not break the pocket affinity ceiling. YELLOW-band chemistry tightens on tetrazole/CONHOH acid bioisosteres + 3-amino-benzofuran-2-COOH core + naphthyl/biphenyl distal aryls; Phase 4h tafamidis-class seeds contribute only 1 of top-20. A tier held.
Problem
Phase 3c v3b (2026-04-24 morning, 12,253 ligands) produced 29 YELLOW with a hard ceiling at mean ΔG −7.28 kcal/mol (Kd 4.57 µM, f_PC 0.34) — MILD-MODERATE adjunct threshold crossed (f_PC ≥ 0.30), NORMAL target (f_PC ≥ 0.50) still distant. Phase 3c v4 was designed to test whether a tighter, fragment-grown library on the v3b winner cluster + Phase 4h tafamidis-playbook bioisosteric seeds could push the ceiling further.
The hypothesis: 1,572 curated hybrids (3 locked cores × 24 distal aryls × 6 ring subs × 4 acid bioisosteres + 22 explicit Phase 4h seeds) would sample the local chemistry neighbourhood densely enough to find a +0.5 to −1.5 kcal/mol gain over v3b, potentially reaching GREEN (f_PC ≥ 0.50).
Method
- Script:
[[pharmacochaperone_phase3c_v4_ensemble_dock.py]](PID 81208, launched 16:47 local, delivered 18:19 local — 92 min wall). - Library:
[[pharmacochaperone_phase3c_v4_library.py]](1,572 ligands: 3 cores — 3-amino-benzofuran-2-COOH, 2-amino-benzoxazole-5-COOH, 2-amino-benzothiazole-6-COOH — × 24 Phase 4h polar distal aryls × 6 ring subs × 4 acid bioisosteres, plus 22 explicit Phase 4h seed SMILES including tafamidis parent, iododiflunisal 3CN/3OMe, 3 v3b reference replays). - Pipeline: identical to Phase 3c v3b. Stage 1 — all 1,572 × snap_008 Vina exh 8 × 3 modes (global-rep screen). Stage 2 — top 50 × 5 k-means-selected Phase 5a MD conformers (snapshots 8/6/15/17/12) Vina exh 16 × 5 modes.
- Aggregation: mean ΔG across 5 snapshots → Kd from ΔG (RT 0.5922 kcal/mol at 310 K); f_PC = bound fraction at [L] = 10 µM × 0.5 rescue efficacy.
- Classification gates (inherited from v3b): GREEN f_PC ≥ 0.50, YELLOW 0.25 ≤ f_PC < 0.50, RED < 0.25.
Results
Verdict distribution
| band | n | f_PC range |
|---|---|---|
| GREEN | 0 | ≥ 0.50 |
| YELLOW | 32 | 0.25 ≤ f_PC < 0.50 |
| RED (top-50 Stage 2) | 18 | < 0.25 |
Top-5 ligands
| rank | name | mean ΔG (kcal/mol) | std | Kd (µM) | f_PC |
|---|---|---|---|---|---|
| 1 | v4__3-amino-benzofuran-2-COOH__2-naphthyl__CONHOH__-CF3 | −7.27 | 0.71 | 4.65 | 0.341 |
| 2 | v4__3-amino-benzofuran-2-COOH__6-F-2-naphthyl__tetrazole__-F | −7.25 | 0.58 | 4.79 | 0.338 |
| 3 | v4__3-amino-benzofuran-2-COOH__4-F-biphenyl__tetrazole__-Me | −7.25 | 0.69 | 4.84 | 0.337 |
| 4 | v4__3-amino-benzofuran-2-COOH__2-naphthyl__CONHSO2Me__-CF3 | −7.24 | 0.65 | 4.86 | 0.337 |
| 5 | v4__3-amino-benzofuran-2-COOH__4-F-biphenyl__tetrazole__-F | −7.22 | 0.65 | 5.07 | 0.332 |
Ceiling comparison v3b → v4
| library | size | top mean ΔG | top Kd (µM) | top f_PC | n GREEN | n YELLOW |
|---|---|---|---|---|---|---|
| Phase 3c v3b | 12,253 | −7.28 | 4.57 | 0.343 | 0 | 29 |
| Phase 3c v4 | 1,572 | −7.27 | 4.65 | 0.341 | 0 | 32 |
Δ ceiling = 0.01 kcal/mol (within Vina noise). Fragment-grow delivered YELLOW-band tightening (more ligands in the 0.25–0.34 f_PC band) but no GREEN.
Chemistry signature of top-20
- Core: 3-amino-benzofuran-2-COOH dominates top-5 (5/5), top-20 (~14/20). 2-amino-benzoxazole-5-COOH and 2-amino-benzothiazole-6-COOH cores (Phase 4h tafamidis-class) do not compete at the top on Vina alone.
- Acid bioisostere (top-20 counts): tetrazole 9, CONHOH 6, CONHSO2Me 4, COOH 0 (naked carboxylate does not appear in top-20).
- Distal aryl: naphthyl / 4-F-biphenyl / biphenyl variants dominate.
- Sources: 19 of top-20 combinatorial; 1 Phase 4h playbook seed. Bioisosteric seeds (tafamidis parent etc.) do not out-dock the combinatorial space.
Verdict
- Outcome: RED-band absolute (no GREEN); YELLOW-band enrichment vs v3b (32 vs 29 hits).
- Key numbers: top f_PC 0.341 (MILD-MODERATE adjunct, NORMAL target unreached); ceiling ΔG flat vs v3b (−7.27 vs −7.28).
- Interpretation: The K1141 pocket has a hard affinity ceiling near mean ΔG ~ −7.3 kcal/mol for the bicyclic-COOH / aryl / acid-bioisostere chemistry class. Fragment-growing within the cluster does not break the ceiling — local chemistry has been saturated. Phase 4h bioisosteric seeds (tafamidis / benzoxazole / benzothiazole cores) do not out-dock the 3-amino-benzofuran-2-COOH combinatorial on Vina — they may still hold wet-lab ADME / off-target advantages, but the mean-ΔG-on-K1141 axis is not theirs.
- Limitations: Vina ensemble mean ΔG is a semi-empirical score with ~0.5 kcal/mol noise on paired comparisons. The apparent ceiling may be an artefact of (a) static-pocket snapshot sampling — all 5 Phase 5a snapshots are WT-Ultra-Mini at 2 ns equilibration, possibly not sampling the productive binding conformation; (b) Vina scoring function systematically underweighting ionic K1141-COOH interactions (salt-bridge geometry is notoriously hard for force-field-approximation docking scorers). The ceiling is in Vina-space, not necessarily in true-ΔG space.
Implications for the K1141 pocket chemistry axis
- Further combinatorial fragment-grow on bicyclic-COOH scaffolds is exhausted. Expected marginal gain per round < 0.5 kcal/mol; library scale would need to move an order of magnitude (~15k+ ligands) for a chance at a lucky tail hit.
- Affinity breakthrough, if it exists, lives outside local chemistry. Options: (a) reversible covalent warheads on Lys ε-NH3+ (Phase 6b queued — would deliver +2–3 kcal/mol if productive); (b) cryptic-pocket discovery on Phase 5d mutant trajectory (Phase 5c on WT snapshot showed no cryptic pocket, but mutant may open one); (c) rational structure-based growth using MD-resolved binding site — requires ligand-bound MD (Phase 5g, not yet run) to identify under-filled pocket regions.
- Static-Vina-ceiling vs true-ΔG-ceiling is an open question. Phase 5h MM-PBSA on top-3 v4 winners (on ligand-bound trajectories from Phase 5g) would test whether the ceiling is physical or a Vina artefact. Until then, the −7.3 kcal/mol ceiling is a scoring-function statement, not a biophysical one.
Ranking delta
- Hypothesis h01: tier A held | mech 3 held | deliv 4 held | misha_fit 4 held
- Next step update: Phase 3c v4 fragment-grow DELIVERED; 32 YELLOW, 0 GREEN, ceiling flat. Medchem axis on K1141 bicyclic-COOH local chemistry exhausted — next moves are Phase 4c-v3b WT-decoy (running) + Phase 4d K1141A salt-bridge decoy (queued) for mechanism validation; Phase 5g holo-MD + Phase 5h MM-PBSA for ceiling interpretation; Phase 5c-mutant cryptic-pocket scan for pocket-discovery alternative. No wet-lab gates shifted.
- Phase 4h tafamidis-playbook status: seeds did not out-dock combinatorial; medchem value is ADME/off-target (not pocket-affinity). Retain as Stage-1 commercial probes, not as lead nominations.
Connections
[part-of]h01 hub[see-also]STRC h01 Phase 3c v3b + 5d Delivery 2026-04-24[see-also]STRC h01 Phase 4h Tafamidis Playbook Library 2026-04-23[see-also]STRC h01 Phase 4c-v3b WT Decoy on v3b YELLOW 2026-04-24[see-also]STRC AF3 Static Pocket Blindness to Loop Dynamics[source]2021-eberhardt-autodock-vina-1-2