STRC Research

STRC h01 Phase 4c-v3b WT Decoy on v3b YELLOW 2026-04-24

5 min read

STRC h01 Phase 4c-v3b WT Decoy on v3b YELLOW

All 29 v3b YELLOW ligands + 5 Phase 4h Tier-1 commercial seeds prefer WT over E1659A mutant on static-structure Vina (frac_prefer_mut = 0/34). v3b YELLOW mean Δ(WT−mut) = −0.538 kcal/mol, median −0.552; Tier-1 seeds mean −0.090, median −0.066. Legacy Phase 4c (2026-04-21) had mean −0.455 on 5 leads. The WT bias is structural — independent of chemistry class — and the higher-affinity a ligand is, the stronger the bias. This falsifies the static pharmacochaperone claim; the dynamic claim (loop 1642-1651 stabilisation) remains untested and requires Phase 5g holo-MD. See STRC AF3 Static Pocket Blindness to Loop Dynamics for methodological framing.

Problem

The original STRC Pharmacochaperone Phase 4c WT Decoy (2026-04-21) ran WT-vs-E1659A docking on 5 Phase 4b leads + 1 positive + 3 negatives. All 5 leads preferred WT (mean ΔΔG(WT−mut) = −0.455 kcal/mol), and Phase 4c was logged as FAIL. That verdict was interpreted as a fundamental challenge to h01.

Two open questions remained:

  1. Chemistry gap. Does the v3b YELLOW cluster (29 sub-5-µM hits, f_PC 0.25-0.34) — which is a qualitatively different scaffold class from the Phase 4b legacy leads — reverse the WT preference?
  2. Phase 4h Tier-1 gap. Do the Phase 4h tafamidis-playbook bioisosteric seeds (tafamidis parent, iododiflunisal 3CN/3OMe, benzoxazole 3CN/3OMe) show the same static signature?

Both sets were docked against the same WT (job4-wildtype.cif) and E1659A mutant (job3-mutant.cif) AF3 structures used in the original Phase 4c, with identical Vina v1.2.7 settings.

Method

  • Script: [[pharmacochaperone_phase4c_v3b_wt_decoy.py]] (new, 2026-04-24).
  • Receptors: pre-prepped PDBQTs from docking_runs/4c/{wt_full,e1659a_mutant}_chainA.pdbqt (identical to legacy Phase 4c).
  • Boxes: identical to legacy Phase 4c — extracted from the original Vina logs.
    • wt_full centre = (4.526, −7.941, −42.352), 18×18×18 Å
    • e1659a_mutant centre = (15.777, −44.660, 16.866), 18×18×18 Å
  • Ligands: pre-prepped PDBQTs from docking_runs/3c_v3/ligands/*.pdbqt (29 v3b YELLOW by name match, f_PC in [0.25, 0.50) from pharmacochaperone_phase3c_v3b_ensemble_dock.json) + docking_runs/3c_v4/ligands/4h_NN_*.pdbqt (5 Tier-1 commercial seeds + diflunisal/niflumic parents).
  • Vina v1.2.7 (2021-eberhardt-autodock-vina-1-2, PMID 34278794), exhaustiveness 32, num_modes 3, cpu 8 — identical to legacy Phase 4c for direct ΔΔG comparability.
  • Wall time: 13 min 24 s for 72 docks (29 + 7 ligands × 2 targets).

Results

v3b YELLOW (n = 29)

metricvalue
median Δ(WT−mut)−0.552 kcal/mol
mean Δ(WT−mut)−0.538 kcal/mol
frac preferring mutant0/29 (0.00)
strongest WT pref−0.989 kcal/mol (nc__3-amino-benzoic__4-F-biphenyl__tetrazole__-CF3)
weakest WT pref−0.162 kcal/mol (nc__2-amino-quinoline-3__biphenyl__CONHOH__-H)

Phase 4h Tier-1 commercial seeds (n = 5)

metricvalue
median Δ(WT−mut)−0.066 kcal/mol
mean Δ(WT−mut)−0.090 kcal/mol
frac preferring mutant0/5 (0.00)
closest to neutral−0.020 kcal/mol (4h_12_iododiflu-3OMe)
largest WT pref−0.163 kcal/mol (4h_01_benzox-3CN-COOH)

Comparison with legacy Phase 4c

cohortnmean Δ(WT−mut)frac prefer mut
Phase 4c legacy leads (2026-04-21)5−0.4550/5
Phase 4c-v3b YELLOW29−0.5380/29
Phase 4c-v3b Tier-1 seeds5−0.0900/5

Affinity-vs-bias correlation

Higher-affinity ligands show stronger WT bias:

  • v3b YELLOW: absolute-ΔG range −8.76 to −7.18 kcal/mol (tight binders), Δ bias −0.54 ± 0.20
  • Tier-1 seeds: absolute-ΔG range −6.26 to −5.91 kcal/mol (weak binders), Δ bias −0.09 ± 0.05

The tafamidis-class seeds bind the pocket an order of magnitude weaker than the v3b combinatorial chemistry, and they are ~6× less WT-biased. This is consistent with a static-pocket-geometry origin for the bias (weak binders engage the pocket shallowly and are insensitive to fine geometric differences).

Verdict

  • Outcome: FAIL (0/29 v3b YELLOW prefer mutant; 0/5 Tier-1 seeds prefer mutant).
  • Key numbers: v3b YELLOW frac_prefer_mut = 0/29, median Δ = −0.552; Tier-1 seeds median Δ = −0.066; legacy −0.455.
  • Interpretation: The WT bias is not a chemistry-shortlist artefact — it is a structural property of the AF3-predicted static pockets. Any ligand tight enough to probe pocket geometry prefers WT by ~0.5 kcal/mol. This falsifies the static geometric form of the pharmacochaperone claim.
  • What this does NOT falsify: the dynamic pharmacochaperone claim (E1659A destabilises loop 1642-1651 on ps-ns timescale; compound dampens loop and shifts folding equilibrium). Static Vina on AF3 structures cannot test a dynamical claim. See STRC AF3 Static Pocket Blindness to Loop Dynamics.
  • Limitations: (1) AF3-mean-structure dependence; (2) Vina force-field known to underweight electrostatic / salt-bridge contributions (ionic interactions score weaker than true ΔG — particularly relevant for a Lys-salt-bridge claim like K1141); (3) no dynamics sampled.

What this changes in the h01 narrative

  1. Phase 4c “FAIL” was not a chemistry problem. The legacy interpretation that Phase 4c failed because the scaffold didn’t engage E1659A pocket was wrong. The static pockets simply don’t discriminate — no chemistry will reverse that.
  2. Higher absolute affinity → stronger WT bias. This trade-off is sharp. Tafamidis-class (weak binders) are essentially neutral; v3b combinatorial (tight binders) are WT-biased by ~0.5 kcal/mol. Medchem optimisation pushing toward GREEN (f_PC ≥ 0.50) will likely increase the WT bias further on static structures.
  3. Medchem selectivity question is orthogonal to affinity question. The chaperone claim lives in dynamics; the affinity claim lives in Vina. They are two different experiments.

Ranking delta

  • Hypothesis h01: tier A held | mech 3 → 2 (static K1141-pocket chaperone claim falsified in Vina force field; dynamic claim remains untested) | deliv 4 held | misha_fit 4 held
  • Next step change: from “Phase 3c v4 fragment-grow (delivered RED, ceiling flat) → wet-lab triage” to “Phase 5g holo-MD on top-3 v3b/v4 leads + E1659A mutant — the only honest test of the dynamic chaperone claim; Phase 5c-mutant cryptic-pocket scan on Phase 5d trajectory; Phase 5h MM-PBSA on holo trajectories to resolve Vina-static-ceiling vs true-ΔG ceiling.” No wet-lab advance.
  • Historical reinterpretation: legacy STRC Pharmacochaperone Phase 4c WT Decoy FAIL verdict reframed as static-docking methodological limit, not evidence against h01 mechanism.

Connections

Graph View

Backlinks