STRC Research

STRC Morning Report 2026-04-25

7 min read

STRC Morning Report — 2026-04-25

Doброе утро. ~7 hours of autonomous overnight work delivered: matched-ensemble pharmacochaperone specificity proof at p < 10⁻¹¹ (mech 3 → 4), 384-ligand v5.2 library with 14-compound ADMET-clean shortlist, off-target panel methodology limit documented, paper draft outline updated. Top wet-lab-ready candidate: v5.2__aq3__adamantyl__CONHOH__-Cl (combined −8.03 kcal/mol, logP 1.94, ADMET flag-free on all 10 gates).

TL;DR (3 lines)

  1. Mech 3 → 4 locked: Phase 5d-WT MD + Phase 5k-WT matched-ensemble APBS gives Δ(MUT − WT) = +4.53 ± 0.46 kT/e = −2.79 ± 0.28 kcal/mol formal-anion preference for E1659A pocket; Welch p = 6.9 × 10⁻¹², Cohen’s d = 3.10. Phase 5j static (+4.37) corrected 1.57× downward — true effect is honest paper-publishable.
  2. v5.2 library expansion (60 → 384 ligands; 9 tails × 5 heads × 7 R3) confirmed adamantyl__CONHOH__-Cl as the wet-lab champion: top combined −8.03 kcal/mol AND ADMET-clean on all 10 gates AND surrounded by 13 structural cousins also clean.
  3. Off-target panel via Vina is uninterpretable — same vdW-dominated blind spot Phase 5j/4i/5k documented. ADMET-AI + wet-lab patch-clamp/enzyme assays remain the operative safety instruments.

Ranking state (before → after overnight)

metric2026-04-24 evening2026-04-25 morning
tierAA
mech3 (per-pose APBS confirmed)4 (matched-ensemble Phase 5k-WT, p=6.9e-12)
deliv33 (held; upgrade-path solidified by 14-compound ADMET-clean shortlist)
misha_fit44

Night’s phases (verdict per phase)

Phase 5d-WT MD (Wakeup 2)

  • 2 ns full-length WT STRC, AMBER14SB/TIP3P/OpenCL, 643k atoms, 14043 s wall (3h54m at 12.98 ns/day)
  • 20 snapshots saved at 100 ps intervals
  • Verified K1141=LYS, E1659=GLU at structure level
  • → artifacts/phase5d_wt_snapshots/ (symlinked from legacy Brain path)

Phase 5k-WT Ensemble APBS (Wakeup 2) — KEY RESULT

  • 20 WT snapshots × APBS nonlinear PBE matched to Phase 5k mutant
  • WT pocket φ = +1.46 ± 1.55 kT/e (16/20 positive; range [−0.81, +4.80])
  • MUT (Phase 5k) +5.99 ± 1.37; Δ = +4.53 ± 0.46 kT/e = −2.79 ± 0.28 kcal/mol formal anion preference
  • Welch t = 9.81, p = 6.9 × 10⁻¹², Cohen’s d = 3.10 (huge effect)
  • Phase 5j static (+4.37 kcal/mol) overestimated by 1.57× — WT static was outlier-pessimistic
  • → STRC h01 Phase 5k-WT Matched Ensemble APBS 2026-04-25

Phase 8d v5.2 Library Generation (Wakeup 2)

  • 384 candidates after drug-likeness filter (MW<500, logP<5, HBA≤10, rotB≤10)
  • 9 tails × 5 acid heads × 7 R3 cross-product
  • 280 non-planar + 104 planar; logP [−0.9, +3.65]; sp3 0.30-0.52
  • Tetrazole head failed kekulization (deferred to v5.3)
  • → artifacts/phase8d_v5_2_library/

Phase 8e v5.2 Dock + Ensemble-APBS Rescore (Wakeup 2)

  • 384/384 docked in 2h15m wall (Meeko + Vina exh=16, parallel-4)
  • Pose-transplant into 20 Phase 5k mutant grids → ensemble-median Δφ
  • Combined score = Vina_ΔG + α × ΔG_formal_ensemble (α=1)
  • Top combined pre-ADMET: biphenyl__CONHSO2Me__-CN at −8.13 (but ADMET-flagged)
  • → artifacts/phase8e_v5_2_dock_rescore/

Phase 8f v5.2 ADMET-AI Triage (Wakeup 2)

  • 30 → 15 ADMET-clean (50%) via 10-gate DrugBank-percentile filter
  • 14-compound shortlist (combined < −7.0 AND clean): 8 adamantyl + 4 1-indanyl + 2 4-F-biphenyl × 4 acid-head classes
  • v5 v1 champion adamantyl__CONHOH__-Cl confirmed #1 ADMET-clean
  • Naphthyl-class DILI-fail rate 7/7 (universal across the library)
  • → STRC h01 Phase 8d 8e 8f v5.2 Library Design 2026-04-25

Phase 8g v5.2 Off-Target Panel (Wakeup 3)

  • 6 ligands × 7 targets = 42 Vina docks (hERG/TMEM16A/KCNQ4/Cx50/TRPM4/COX-1/COX-2)
  • 0/6 pass 100× selectivity including v3b reference
  • Verdict: methodological caveat — same Vina vdW-blindness as on-target case
  • One genuine positive: 1-indanyl__COOH__-F rejects COX-2 (ΔG = +1.25 kcal/mol = no binding)
  • ADMET-AI + Phase 6c remain operative safety signals; wet-lab is decisive
  • → STRC h01 Phase 8g v5.2 Off-Target Panel 2026-04-25

Paper draft (Wakeups 2-3)

  • Outline at ~/STRC/notes/STRC Paper Draft Outline 2026-04-25.md
  • Abstract updated with Phase 5k-WT numbers
  • §1 Intro / §2.1-§2.6 Results / §3 Discussion / §4 Methods sections all sketched
  • 6 figures placeholdered
  • 8 outstanding-before-submission items listed (wet-lab validation, AF3 Phase 7H, α sensitivity, etc.)

Top wet-lab-ready candidates

Wet-lab synthesis priority list (5 compounds across 3 scaffold classes):

#namecombined (kcal)VinaAPBSADMET flagslogPMWhERG %DILI %scaffold
1v5.2__aq3__adamantyl__CONHOH__-Cl−8.03−6.68−1.350/101.94371.98073adamantyl + CONHOH
2v5.2__aq3__adamantyl__CONHOMe__-Cl−7.74−6.59−1.160/102.11385.97675adamantyl + CONHOMe
3v5.2__aq3__1-indanyl__COOH__-Me−7.58−6.11−1.470/102.31318.463721-indanyl + COOH
4v5.2__aq3__adamantyl__CONHSO2Me__-F−7.53−6.70−0.830/101.24417.56682adamantyl + CONHSO2Me
5v5.2__aq3__1-indanyl__COOH__-F−7.47−6.04−1.420/102.22322.360751-indanyl + COOH

All 5 pass 10-gate ADMET filter; off-target Vina selectivity is uninterpretable per Phase 8g methodology limitation.

Recommended primary synthesis target: #1 adamantyl__CONHOH__-Cl — first-in-class ADMET-clean pharmacochaperone for STRC E1659A. Recommended backup #1: #3 1-indanyl__COOH__-Me — different scaffold (sp2 indane + COOH) for orthogonal medchem. Recommended backup #2: #4 adamantyl__CONHSO2Me__-F — same adamantyl tail with sulfonamide head; lowest logP (1.24) for highest aqueous solubility.

Open questions for daytime session with Egor

  1. Wet-lab synthesis decision — pursue top-3 simultaneously or only #1? (Adamantyl + 1-indanyl scaffolds are orthogonal; risk-mitigated portfolio approach favors top-3.)
  2. AF3 Phase 7H submission — 12 jobs queued at ~/STRC/models/af3_jobs_2026-04-24_phase7h/ blocked on Google auth. ~5 min of your time to submit; ~6-12h AF3 compute. Adds orthogonal cross-check on 6 NSAID-class parents (CCD-format, can’t submit v5.2 SMILES).
  3. v5.3 design questions:
    • Resurrect tetrazole head with explicit kekulization in SMILES template?
    • 2-amino-quinoxaline (2 N’s in core ring) as alternative scaffold?
    • 1-adamantyl-2-amino variants (regiochemistry change)?
  4. Phase 8h APBS off-target (~15h compute) — if you want a “fair” Coulomb-based off-target selectivity to replace the Vina-uninterpretable Phase 8g, this is the way. Expensive but clean for paper.
  5. Wet-lab outsourcing strategy — decide on contract lab for thermal shift / SPR Kd / patch-clamp hERG / Caco-2. Budget ~$15k / ~6 weeks per Phase 8g recommendation.
  6. Paper venue — Nature Chem Biol (mechanism-heavy, fits current data) vs J Med Chem (medchem-heavy, would benefit from wet-lab Kd data). Recommend NCB if wet-lab takes >3 months; JMC if <3 months.

Paper draft status

~/STRC/notes/STRC Paper Draft Outline 2026-04-25.md:

  • Abstract — written; matched-ensemble numbers + 14-compound shortlist included
  • §1 Intro — outlined (DFNB16 background, pharmacochaperone precedent, electrostatic asymmetry hypothesis)
  • §2 Results — outlined for 6 sub-sections, all keyed to existing phase proofs
  • §3 Discussion — outlined (Vina blindness, design principle, generalizability, limitations)
  • §4 Methods — outlined; references all scripts in scripts inventory
  • §5 Figures — 6 placeholdered: pocket isosurface, Vina vs APBS scatter, pose-transplant, ensemble distribution, top-5 v5 structures
  • §6 Outstanding — 8 items: wet-lab validation, AF3 Phase 7H, α sensitivity, blinded retrospective on tafamidis/lumacaftor, etc.

Status: ready for prose-writing pass. Estimated time to first draft: 2-3 days of writing if all data stays as-is.

Sanity check

  • Wikilink balance: all [[...]] opens have matching ]] closes (grep mismatch check passed)
  • Hub frontmatter: tier A | mech 4 | deliv 3 | misha_fit 4 | excerpt updated for matched-ensemble result
  • Log: all overnight phases prepended (Phase 8c → 5d-WT/5k-WT/8d/8e/8f → 8g)
  • Scripts inventory: all 8 new scripts (5d-WT, 5k-WT, 8c, 8d, 8e, 8f, 8g, plus support) registered
  • 14 phase proof files dated 2026-04-24/25 in ~/STRC/phases/
  • All artifacts written to ~/STRC/models/artifacts/ (Phase 5d-WT snapshots symlinked from legacy Brain path)

Compute summary (overnight)

  • Phase 5d-WT MD: 14043 s wall (3h54m), Metal OpenCL, 643k atoms
  • Phase 5k-WT APBS: ~25 min, parallel-4, 20 snapshots
  • Phase 8d library gen: ~10 s
  • Phase 8e dock: 2h15m, parallel-4 Vina, 384 ligands
  • Phase 8c ADMET (live, before cron): ~1 min
  • Phase 8f ADMET (v5.2): ~1 min
  • Phase 8g off-target: ~25 min, parallel-4, 42 docks
  • Total compute wall: ~7h matching the autonomous run window

Connections

⚠ Checklist for Egor (morning, ~5-15 min total)

  1. Read TL;DR + Top wet-lab candidates section above (1 min)
  2. Decide primary synthesis target: #1 alone, top-3 portfolio, or different choice (2 min)
  3. AF3 Phase 7H Google auth submission if you want the orthogonal NSAID-parents cross-check (~5 min in browser; jobs at ~/STRC/models/af3_jobs_2026-04-24_phase7h/)
  4. Skim Phase 5k-WT proof for the matched-ensemble specificity number (3 min)
  5. Review Paper Draft Outline abstract + §2.4 (matched ensembles) — that’s the paper’s central quantitative claim now (5 min)
  6. Decide on wet-lab contract — patch-clamp + thermal shift on top-3 (decision; the actual outsourcing can wait)
  7. Optional: open question on v5.3 design direction (tetrazole resurrection, quinoxaline scaffold, 1-adamantyl regio change) — to-do or defer

If everything checks out, the night was a clean delivery. Если что-то не сходится — есть bullet pointing the right phase proof to drill into.

Graph View

Backlinks