STRC Pharmacochaperone Phase 4a — K1141 pocket reproduces across four of five AF3 STRC CIFs including Ultra-Mini × TMEM145

The K1141-adjacent loop-facing druggable subpocket found in WT (Phase 2B, druggability 0.86) is not a one-job artifact. It reproduces in 4 of 5 independently folded AF3 STRC structures with druggability 0.82–0.92, K1141 in its lining residues in every case, and centroid ≤8.5 Å from K1141 Cα. Gate PASS → Phase 4b docking authorised.

Method

Applied the Phase 2B ray-casting + depth-peak scorer unchanged (18 Å scan radius around loop-1642-1651 centroid, 0.8 Å grid, 9/14 rays required for burial, 50-voxel minimum cluster) to five AF3 CIFs. Residue numbering is remapped to each CIF’s local frame because AF3 renumbers truncated constructs from 1. Offsets (real STRC residue = CIF residue + offset):

  • job4-wildtype.cif — 0
  • job3-mutant.cif — 0
  • job5-mini-strc.cif — +593 (mini 594-1775)
  • job-h-strc-cterm-only.cif — +1074 (Ultra-Mini proxy 1075-1775)
  • job-ultramini-x-tmem145-full.cif — +1074 (clinical Ultra-Mini + TMEM145)

Absolute Phase 2B box coordinates (7.7, −5.4, −41.5) Å don’t transfer across CIFs because AF3 rotates complexes arbitrarily. Replaced with a reference-frame-free test: best K1141-inclusive subpocket must have (a) druggability ≥0.70, (b) K1141 in lining residues, (c) centroid ≤12 Å from local K1141 Cα, (d) centroid ≤12 Å from local loop-1642-1651 centroid. Gate: pass in ≥4/5 CIFs.

Result

CIFConstructDruggabilityVolume (ų)K1141 in pocketd(K1141 Cα)d(loop centroid)Pass
job4-wildtype.cifWT 1-17750.8698.87.42 Å9.30 Å
job3-mutant.cifE1659A 1-17750.8270.14.79 Å9.68 Å
job5-mini-strc.cifmini 594-17750.92145.98.46 Å17.05 Å
job-h-strc-cterm-only.cifUltra-Mini proxy 1075-17750.90129.06.85 Å10.00 Å
job-ultramini-x-tmem145-full.cifUltra-Mini + TMEM1450.90117.87.45 Å8.65 Å

Verdict: 4/5 PASS — gate cleared.

Interpretation

  1. Pocket is load-bearing structure, not AF3 noise. Druggability is uniformly high (0.82–0.92, narrow spread) across five independently folded jobs with different sequence lengths, different complex partners, and different coordinate frames. A one-job artifact would show high variance or vanish in the truncations.
  2. Mutant preserves the pocket. E1659A shows 0.82 druggability, higher K1141 proximity (4.79 Å vs 7.42 Å in WT — consistent with Phase 1B’s observation that K1141 shifts 3.4 Å when E→A is introduced). The pocket is intact in the mutant. Rescuing ligands have a pocket to bind to.
  3. Clinical Ultra-Mini × TMEM145 complex carries the pocket. The 700-bp Ultra-Mini construct that will actually be delivered by AAV (drug-rescueable maternal allele co-expressed with paternal null is the clinical scenario) retains the pocket in the presence of the bound TMEM145 partner — druggability 0.90, centroid 8.65 Å from loop, 7.45 Å from K1141. This is the single most important row in the table: it means the Phase 4b-4f screen can run directly on the clinical construct CIF and the result will transfer to the treatment scenario.
  4. Mini 594-1775 fails the loop-proximity gate, not the druggability gate. Its K1141-inclusive subpocket has druggability 0.92 but is 17 Å from the loop centroid — a different geometry, likely because the extra N-terminal residues (594-1074) in the mini construct reorganise the LRR face. Not used as docking target for Phase 4b; Ultra-Mini × TMEM145 is preferred anyway.

Files / Models

  • ~/STRC/models/pharmacochaperone_phase4a_pocket_reproducibility.py — driver (reuses Phase 2B scorer verbatim, iterates over 5 CIFs)
  • ~/STRC/models/pharmacochaperone_phase4a_pocket_reproducibility.json — full per-CIF output (all subpockets, lining residues in real STRC numbering, geometry)

Gate status for Phase 4 pipeline

  • 4a — Pocket reproducibility — PASS ← this proof
  • 4b — Vina + GNINA real docking — next
  • 4c — WT control docking — next
  • 4d — K1141A in-silico decoy — next
  • 4e — Off-target selectivity scan — next
  • 4f — Interface-rescue MM-GBSA — next
  • 5 — GROMACS MD — next

Phase 4b docking target selected: job-ultramini-x-tmem145-full.cif (clinical construct, complex-stabilised pocket, druggability 0.90).

Ranking delta

  • STRC Pharmacochaperone Virtual Screen E1659A: no tier change. Stays S-tier. Evidence depth +1 — Phase 0 gate strengthened from “druggable pocket in WT (n=1 CIF)” to “druggable pocket in WT + mutant + Ultra-Mini solo + Ultra-Mini × TMEM145 complex (n=4/5 CIFs)“. Mechanism axis no longer vulnerable to the “single-CIF artifact” kill criterion. Next step column in STRC Hypothesis Ranking updated from “run Phase 4a pocket reproducibility gate” → “run Phase 4b Vina + GNINA real dock on Ultra-Mini × TMEM145 CIF”.
  • STRC Pharmacochaperone K1141 Fragment Pocket: no change. Phase 2B finding (WT only) now has 3 additional independent confirmations in mutant + two truncations + TMEM145 complex.
  • STRC Mini-STRC Single-Vector Hypothesis: no change. Consumed the Ultra-Mini × TMEM145 CIF as a docking target; did not modify Mini-STRC evidence.
  • All other S/A/B/C hypotheses: no change.

Connections