STRC h01 Phase 4d K1141A Double-Mutant Decoy

Truncating LYS1141 → ALA on the E1659A background (double mutant) and re-docking 29 v3b YELLOW + 7 Phase 4h Tier-1 / control seeds delivers a mean binding loss of only 0.18 kcal/mol (median 0.26, max 0.65, n=29). 0 of 34 acidic ligands lose ≥ 2 kcal/mol — the gate threshold for a load-bearing salt-bridge anchor. 7 of 36 ligands actually dock BETTER on K1141A (largest gain −0.79 kcal/mol). The K1141 ε-NH3+ is not the load-bearing anchor that the h01 pharmacophore (STRC Pharmacochaperone K1141 Fragment Pocket) claimed.

Problem

The h01 pharmacochaperone scaffold rationale — built from STRC Pharmacochaperone K1141 Fragment Pocket and STRC Pharmacophore Model K1141 Pocket — centres on a salt bridge between an acidic warhead (COOH / tetrazole / CONHOH) on the ligand and the K1141 ε-NH3+ on the receptor. Every Phase 3c v2/v3/v3b/v4 library was designed around this anchor. If the anchor is load-bearing, removing K1141 ε-NH3+ should cost ≥ 2 kcal/mol for acidic ligands.

The original STRC Pharmacochaperone Phase 4 Plan Phase 4d scaffold was written 2026-04-21 but never run — STRC Computational Scripts Inventory flagged it **scaffold** (skipped per Phase 4c interpretation). After STRC AF3 Static Pocket Blindness to Loop Dynamics reframed Phase 4c as a methodological limit (not a falsification), the K1141 test became the falsifiable sub-claim on the critical path.

Method

  • Script: [[pharmacochaperone_phase4d_k1141a_double_mutant_dock.py]] (new 2026-04-24, supersedes unrun pharmacochaperone_phase4d_k1141a_decoy.py scaffold which pointed at the Ultra-Mini × TMEM145 construct rather than the full-length E1659A disease target).
  • Receptor mutagenesis (pure Python, no pymol/modeller dependency): load docking_runs/4c/e1659a_mutant_chainA.pdb → for residue A1141, keep N/CA/C/O/CB atoms, drop CG/CD/CE/NZ, rename LYS → ALA → docking_runs/4d/e1659a_k1141a_chainA.pdb. No distal repacking. obabel --partialcharge gasteiger -p 7.4 -xr → PDBQT.
  • Box: identical to Phase 4c e1659a_mutant, centre (15.777, −44.660, 16.866), 18 × 18 × 18 Å. ALA sidechain volume is ~45 ų smaller than LYS, but the backbone is unchanged — pocket centre drift is sub-1 Å.
  • Ligands: 29 v3b YELLOW + 5 Phase 4h Tier-1 commercial seeds + diflunisal/niflumic parents as controls; identical pre-prepped PDBQTs as Phase 4c-v3b.
  • Vina v1.2.7, exhaustiveness 32, num_modes 3, cpu 8. Paired against Phase 4c-v3b e1659a_mutant_dG (same receptor before sidechain truncation) — no re-run of intact-K1141 arm needed.
  • Per-ligand quantity: salt_bridge_loss_kcal = ΔG_K1141A − ΔG_E1659A_K1141_intact (positive = binding lost on K1141 removal, i.e., the salt bridge was load-bearing).
  • Gate: PASS if ≥ 70 % of v3b YELLOW show loss ≥ 2 kcal/mol.

Results

v3b YELLOW (n = 29)

metricvalue
median loss+0.257 kcal/mol
mean loss+0.176 kcal/mol
frac loss ≥ 2 kcal/mol0/29 (0.00)
frac loss ≥ 1 kcal/mol0/29 (0.00)
frac loss ≥ 0.5 kcal/mol4/29 (0.14)
largest loss+0.647 (nc__2-amino-quinoline-3__3_5-diMe-phenyl__tetrazole__-F)
largest negative loss (K1141A binds BETTER)−0.792 (nc__2-amino-quinoline-3__1-naphthyl__CONHOH__-CF3)

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

metricvalue
median loss+0.139 kcal/mol
mean loss+0.143 kcal/mol
frac loss ≥ 2 kcal/mol0/5 (0.00)
largest loss+0.294 (4h_05_tafamidis_parent)
smallest loss+0.005 (4h_02_benzox-3OMe-COOH)

Full 36-ligand breakdown

classvalue
positive loss (K1141A loses binding)29/36
negative loss (K1141A binds BETTER)7/36
loss ≥ 0.5 kcal/mol4/36
loss ≥ 1.0 kcal/mol0/36
loss ≥ 2.0 kcal/mol0/36
niflumic parent loss+0.397
diflunisal parent loss−0.053

Top-5 losses (largest K1141 dependence)

nameintact (mut)K1141Aloss
nc__2-amino-quinoline-3__3_5-diMe-phenyl__tetrazole__-F−7.145−6.498+0.647
nc__2-amino-quinoline-3__3-CF3-phenyl__CONHSO2Me__-CF3−7.338−6.707+0.631
nc__3-amino-benzofuran-2-COOH__4-F-biphenyl__tetrazole__-Me−7.650−7.102+0.548
nc__3-amino-benzofuran-2-COOH__4-F-biphenyl__CONHSO2Me__-F−7.186−6.666+0.520
nc__2-amino-quinoline-3__4-F-biphenyl__tetrazole__-Me−7.296−6.802+0.494

Bottom-5 losses (K1141A BINDS BETTER)

nameintact (mut)K1141Aloss
nc__2-amino-quinoline-3__1-naphthyl__CONHOH__-CF3−7.298−8.090−0.792
nc__2-amino-quinoline-3__biphenyl__CONHOH__-H−7.014−7.423−0.409
nc__3-amino-benzofuran-2-COOH__4-F-biphenyl__CONHOH__-H−7.029−7.339−0.310
nc__3-amino-benzofuran-2-COOH__1-naphthyl__CONHOH__-CF3−7.269−7.525−0.256
nc__2-amino-quinoline-3__3_5-diMe-phenyl__tetrazole__-CF3−7.221−7.286−0.065

Verdict

  • Outcome: FAIL (0/29 v3b YELLOW hit the 2 kcal/mol salt-bridge gate).
  • Key numbers: v3b YELLOW mean loss +0.18, median +0.26, max +0.65; Tier-1 mean +0.14. 7/36 ligands dock BETTER on K1141A (max gain −0.79).
  • Interpretation: The K1141 ε-NH3+ is not the load-bearing anchor for v3b / Phase 4h acidic ligands at the static-Vina level. Removing the Lys sidechain costs, on average, a quarter kcal/mol — an order of magnitude less than the ~2 kcal/mol expected for a bona fide ionic H-bond. Some CONHOH-class ligands actually gain binding on K1141A (negative loss), suggesting the intact Lys sidechain is sterically interfering with their preferred binding mode. The scaffold direction was rationalised around a pharmacophore that is not load-bearing in the docking model.
  • Limitations:
    1. Vina force field systematically underweights Coulombic / salt-bridge contributions; the true ΔG cost of K1141 removal is likely higher than the Vina-measured +0.18 kcal/mol. An MM-PBSA or FEP+ analysis on ligand-bound trajectories of WT-K1141 vs K1141A would give a force-field-independent answer.
    2. No distal repacking — the ALA sidechain leaves a ~45 ų cavity next to CB. Ligands that reorient into that cavity may over-estimate “binding gain” (the 7 negative-loss rows above). A short local minimisation before docking would clean this up.
    3. Only static-structure test — if the dynamic chaperone claim is correct, the K1141-COOH contact may form transiently in ligand-bound MD and be invisible to single-snapshot docking.
  • What this validates: the decoy concept itself — the K1141A double mutant docks at reasonable ΔG (−6.5 to −8.1 kcal/mol range, comparable to intact) without structural artifacts. The mutagenesis protocol (sidechain truncation without repacking) is clean.

What this changes in the h01 narrative

  1. “K1141 salt-bridge pharmacophore” is a Vina-unsupported hypothesis. Removing the salt-bridge partner changes binding by less than the Vina noise floor (±0.5 kcal/mol paired). This does not categorically falsify the claim — the Vina scoring function’s electrostatic treatment is the limiting factor — but it removes static-Vina as evidence for the claim.
  2. Medchem constraint loosens. The scaffold rationale that every lead must carry an acidic warhead (COOH / tetrazole / CONHOH / CONHSO2Me) to engage K1141 is unsupported by this test. Neutral / amide / alcohol analogs become admissible if they bind the pocket equally well by other mechanisms. Follow-up: re-screen a subset of v4 library replacing the acid with neutral isosteres.
  3. Salt-bridge confirmation now requires MM-PBSA or FEP+. Vina cannot answer this question. The Phase 5g holo-MD + Phase 5h MM-PBSA plan should include a K1141A control arm (top-3 ligands × {WT-K1141, K1141A}) to quantify the true salt-bridge ΔΔG.

Ranking delta

  • Hypothesis h01: tier A held | mech 2 held (already reduced by Phase 4c-v3b; Phase 4d provides independent evidence that the K1141 static-anchor claim was over-stated) | deliv 4 held | misha_fit 4 held
  • Next step change: add “K1141A control arm to Phase 5g / Phase 5h plan” explicitly; deprioritise naked-COOH scaffold nomination; keep CONHOH/tetrazole tracks open (they show the steric-interference pattern, not a salt-bridge pattern). No wet-lab advance.
  • Scripts inventory: mark pharmacochaperone_phase4d_k1141a_decoy.py as superseded (pointed at wrong construct); pharmacochaperone_phase4d_k1141a_double_mutant_dock.py is the active replacement, run 2026-04-24.

Connections