STRC h01 Phase 9d QM-MM Pocket 2026-04-27 SKELETON

STRC h01 Phase 9d — QM/MM Pocket Energetics (SKELETON)

Placeholder. Quantum-mechanical treatment of the binding pocket and ligand to test whether the WT-vs-mut energetic differential survives when classical-charge approximations are removed. No production calculation yet.

Question

Treating the binding-pocket residues + ligand quantum-mechanically and the rest of the protein + solvent classically:

1. Does the WT-vs-mutant interaction-energy differential agree in sign and magnitude with the MM-based pipeline (Phases 5e/5j/9b)?
2. Are there nontrivial polarization, charge-transfer, or hydrogen-bond directionality contributions that the MM force field cannot capture?
3. Does the result persist across QM functionals (B3LYP-D3 vs ωB97X-D vs MP2)?

Method

Two staged calculations:

• Stage 1 — QM/MM single point on representative MD frames (from Phase 5d/9b).
  • QM region: ligand + side chains within 5 Å of ligand heavy atoms (typically 6-12 residues including E1659/A1659 and K1141 if proximal).
  • MM region: rest of protein + waters + ions.
  • Functional: B3LYP-D3(BJ) / 6-31G(d,p) for screening; ωB97X-D / def2-TZVP for production.
  • Embedding: electrostatic embedding (point-charge polarization of QM density).
• Stage 2 — QM/MM constrained optimization of bound complex (with QM region relaxed, MM frozen beyond a 2-shell flexible region).
  • Reports ΔE_int(QM/MM) per system.

Tooling candidates: ORCA + AMBER (best ratio of features to setup pain), CP2K (built-in QM/MM, fast), Q-Chem + AMBER, ChemShell (most flexible). NAMD-QwikMD QM/MM is a usable GUI route.

Why orthogonal

• vs all MM-based phases: QM captures polarization (induced dipoles), partial charge transfer between ligand and protein, hydrogen-bond cooperativity, and proper directionality of E1659 carboxylate vs K1141 ammonium interactions — none of which fixed-point-charge MM force fields represent correctly.
• Especially relevant for h01: the proposed mechanism is electrostatic (charge-charge salt bridge between K1141 and ligand carboxylate, with E1659 nearby). MM models charges as fixed point monopoles. QM/MM tests whether the electrostatic story holds with proper electron density.
• Disambiguates: if FEP (9a) and MM-PBSA (9b) agree but QM/MM disagrees → MM force field is artifactually inflating the differential; honest interpretation must drop deliv tier. If all four agree → mechanism is robust at multiple physics levels.

Inputs needed

• Representative bound poses from Phase 5e/5q (top-3 ligands × {WT, mutant}).
• Equilibrated MD snapshots from Phase 5d/9b (for ensemble averaging if budget permits).
• AMBER topology and trajectory files.
• Hardware: CPU-heavy. ωB97X-D/def2-TZVP single point on ~150-atom QM region: minutes to hours per frame on 16-32 cores. ChemShell or CP2K on cluster.
• Caution: QM/MM is the slowest of the 5 tracks. Plan accordingly.

Smoke test (1-day, theoretical)

Single bound pose × 1 variant × QM region of ligand + 5 closest residues:

1. Snap one frame from Phase 5d production trajectory.
2. Define QM region (ligand + 5 nearest residue side chains, neutral truncation at Cα-Cβ via link atoms).
3. B3LYP-D3 / 6-31G(d) single-point with electrostatic embedding.
4. Check: SCF converges; total QM energy is reasonable (within range of small-protein-cluster QM literature); MM-derived background does not blow up the QM Hamiltonian.

Smoke pass: SCF converges in <50 cycles, no negative eigenvalues, Mulliken charges on ligand are chemically sensible (e.g., carboxylate O at ~-0.6 e).

Smoke fail: SCF does not converge → QM region truncation is bad (charged residue cut at the wrong place) or basis too small. Redesign QM region.

Production protocol (theoretical)

• 5-10 representative MD frames per ligand × variant.
• ωB97X-D / def2-TZVP single points with electrostatic embedding.
• Constrained QM/MM geometry optimization on 1-2 frames per variant for ΔE_int.
• Energy-decomposition analysis (LMO-EDA or ALMO-EDA in Q-Chem) on optimized geometries → quantify electrostatic vs polarization vs charge-transfer contributions to the differential.
• Output: ΔE_int(WT) vs ΔE_int(mut), ensemble-averaged across frames; EDA breakdown.

Pass criteria

• PRIMARY: sign of ΔE_int(WT) − ΔE_int(mut) matches MM pipeline for ≥1 lead ligand.
• SECONDARY: EDA shows electrostatic + polarization is the dominant differential contribution (≥ 60% of total), confirming the mechanism narrative.
• TERTIARY: result robust across B3LYP-D3 vs ωB97X-D (sign agreement on all functionals tested).
• FAIL state: QM/MM gives opposite sign of differential → MM force field is biased; the K1141/E1659 electrostatic story needs to be reframed in terms of polarization/charge-transfer, or the differential is real but smaller than MM suggests.

Known artifacts and risks

• QM region cutoff: too small (5 Å) misses long-range polarization; too large (12 Å) becomes computationally infeasible. Convergence test on QM-region size is mandatory before production.
• Link atoms at QM/MM boundary: standard hydrogen-link is fine for backbone Cα-Cβ cuts; avoid cutting through aromatic rings or charged side-chain centers.
• Functional choice: B3LYP-D3 systematically underestimates dispersion in some pockets; ωB97X-D and DLPNO-CCSD(T) are better but expensive. Run at least 2 functionals.
• MM polarization back-reaction: standard electrostatic embedding does not let MM atoms polarize back to QM. Polarizable MM (AMOEBA) closes this loop but is heavy. Skip for production; mention as limitation.
• Single-frame bias: QM/MM single-points on a single frame are noisy. Ensemble of 5-10 frames is the minimum honest result.

References (canonical)

• Warshel, Levitt 1976. J Mol Biol 103:227 — original QM/MM (Nobel 2013).
• Senn, Thiel 2009. Angew Chem Int Ed 48:1198 — QM/MM methods review.
• Lin, Truhlar 2007. Theor Chem Acc 117:185 — link-atom and embedding schemes.
• Khaliullin, Cobar, Lochan, Bell, Head-Gordon 2007. J Phys Chem A 111:8753 — ALMO-EDA.
• Kästner et al. 2009. J Comput Chem 30:2237 — ChemShell QM/MM.

Status

• 2026-04-27 — skeleton created. Software install status: TBD (ORCA likely; AmberTools already present). No runs performed.

Ranking delta

• No change. Skeleton only.

Connections

• [parent] STRC h01 Phase 9 Orthogonal Cross-Checks Plan 2026-04-27
• [see-also] h01 hub
• [see-also] STRC h01 Phase 5j APBS WT vs Mutant Pocket Electrostatics 2026-04-24 — continuum-electrostatics counterpart QM/MM stress-tests
• [see-also] STRC h01 Phase 4i APBS Pose-Transplant Rescore 2026-04-24
• [about] Misha