STRC Research

STRC h01 Phase 9b Holo MD MM-PBSA 2026-04-27 SKELETON

6 min read

STRC h01 Phase 9b — Holo MD + MM-PBSA (SKELETON)

Placeholder. Long-trajectory complex stability + isolated electrostatic free energy contribution as an independent check on the docking-derived differential. No production calculation yet.

Question

Over a multi-hundred-nanosecond holo trajectory:

  1. Does the docked pose remain stable (RMSD plateau, no spontaneous unbinding)?
  2. What is ΔG_bind(MM-PBSA) for mutant vs WT, and does the differential survive?
  3. Decomposed by residue: is the differential electrostatically dominated, and is E1659/K1141 the dominant contributor?

Method

Standard holo MD followed by end-point free-energy rescoring:

  1. Solvate, ionize, equilibrate (NVT then NPT) the docked complex in a periodic box.
  2. Production NPT MD at 300 K, 1 atm, 100-500 ns per replicate, 3 replicates per system.
  3. Sample 200-500 evenly-spaced frames from the back-half of each trajectory.
  4. MM-PBSA (or MM-GBSA) rescoring per frame: ΔG = ΔE_MM + ΔG_solv − TΔS.
  5. Per-residue decomposition (gmx_MMPBSA decomposition: True) to identify dominant contributors.

Tooling: GROMACS for MD; gmx_MMPBSA (Valdés-Tresanco 2021) or MMPBSA.py from AmberTools for rescoring.

Why orthogonal

  • vs Vina/Boltz-2 docking: explicit solvent + entropy + long-time-scale conformational sampling instead of single-pose scoring.
  • vs tauRAMD: equilibrium dynamics, no pulling force, gives ΔG not τ.
  • vs APBS continuum: per-frame implicit solvation but on a time-evolving protein-ligand complex (APBS is on a single static frame).
  • Caveat: shares force field with FEP track 9a; if FF is the bias, both inherit it. Disambiguator is 9d (QM/MM).

Inputs needed

  • Phase 5q top-3 to top-5 ligands with parameterised topologies (GAFF2 or OpenFF).
  • WT and mutant docked complexes from Phase 5e v5.3.
  • Force field: AMBER ff19SB / ff14SB protein + OpenFF or GAFF2 ligand + TIP3P or OPC water.
  • Box: 12 Å padding, neutralised with 0.15 M NaCl.
  • Hardware: GPU. M5 Max MPS for GROMACS — verify before launch (Apple Silicon first rule). Plain MD on M5 should work; if GPU memory tight on solvated box, fall back to remote.

Smoke test (1-day, theoretical)

Single ligand × 2 variants (WT, mut) × 1 replica × 10 ns production:

  1. Solvate, equilibrate, run 10 ns NPT.
  2. Compute backbone RMSD and ligand RMSD time-series.
  3. Run gmx_MMPBSA on 50 evenly-spaced frames.

Smoke pass: ligand RMSD < 5 Å in both systems (pose stable enough to extend), MM-PBSA ΔG returns finite numbers with consistent ranking direction at 10 ns.

Smoke fail: ligand spontaneously unbinds in 10 ns (then existing docked pose is not metastable — investigate before extending) or MM-PBSA pipeline numerical failure.

Production protocol (theoretical)

  • Per ligand × variant: 3 replicates × 200 ns each = 600 ns.
  • 4 ligands × 2 variants = 4.8 μs total. ~3-5 days on a single A100 if batched.
  • MM-PBSA on 300 frames per replicate (back half), per-residue decomposition on 100 frames.
  • Report: ΔG ± σ across replicates; per-residue heatmap (W vs M).

Pass criteria

  • PRIMARY: ligand-RMSD plateau < 4 Å in mutant for ≥2 of 3 replicates (complex is metastable on production timescale); divergent behaviour in WT (unbinding or higher RMSD) is supportive.
  • SECONDARY: ΔG(MM-PBSA, mutant) − ΔG(MM-PBSA, WT) < −1 kcal/mol on lead ligand.
  • TERTIARY: per-residue decomposition flags E1659 (or its WT analogue) and/or K1141 as a top-5 contributor by absolute energy.
  • FAIL state: complex unbinds in mutant within 100 ns or per-residue decomposition does not flag the proposed mechanism residues → reconsider mechanism narrative.

Known artifacts and risks

  • MM-PBSA entropy: T·ΔS via normal-mode analysis is noisy and slow; many groups omit it for relative comparisons. Document choice.
  • PB grid spacing / dielectric: gmx_MMPBSA defaults to εsol=80, εint=1; sensitivity to εint is real. Run at εint ∈ {1, 2, 4} as a robustness check.
  • Sampling: 200 ns may be too short for slow conformational rearrangements; use replica-exchange or accelerated MD (GaMD candidate extra) if RMSD doesn’t plateau.
  • Co-alchemical-charge issue is not present here (no λ mutation), but starting structure for mutant comes from in-silico point mutation of the WT (Modeller/Rosetta) — equilibration must absorb any minor steric clashes from the mutation rebuild.

References (canonical)

  • Genheden, Ryde 2015. Expert Opin Drug Discov 10:449 — MM-PBSA review.
  • Wang et al. 2019. Front Mol Biosci 6:165 — MM-PBSA accuracy.
  • Valdés-Tresanco et al. 2021. J Chem Theory Comput 17:6281 — gmx_MMPBSA.
  • Hou et al. 2011. J Chem Inf Model 51:69 — MM-GBSA accuracy.
  • Sun et al. 2014. Phys Chem Chem Phys 16:22035 — MM-PBSA dielectric sensitivity.

Status

  • 2026-04-27 — skeleton created. gmx_MMPBSA install status: TBD. No runs performed.
  • 2026-04-27 — lite residue-field smoke delivered from existing matched WT/MUT MD snapshots. Broad charged-neighbourhood sign supports E1659A > WT (18 A delta +0.181 q/A, p=1.54e-10), but 9 A inner shell opposes. Production holo-MD MM-PBSA remains pending. → STRC h01 Phase 9b-lite Residue Field Decomposition 2026-04-27
  • 2026-04-27 — MM-PBSA tooling preflight delivered. GROMACS, gmx_MMPBSA, MMPBSA.py, cpptraj, tleap, and antechamber are callable, but the direct Phase 9b smoke is blocked by missing matched WT holo input and absent native GROMACS/Amber topologies. Post-hoc Phase 5q OpenMM → Amber conversion is invalid because ParmEd finds 37,405 bonds without Amber bond types. → STRC h01 Phase 9b MM-PBSA Tooling Preflight 2026-04-27
  • 2026-04-27 — matched Amber input compatibility delivered. Builder now creates matched WT/E1659A dry complex/receptor/ligand prmtops and one-frame MMPBSA.py GB reads both sets to completion. The one-frame energies are nonphysical (huge VDWAALS), so production remains pending geometry relaxation and sane-frame smoke. → STRC h01 Phase 9b Matched Amber Input Compatibility 2026-04-27
  • 2026-04-27 — min-relax multi-frame smoke delivered. Repaired matched Amber inputs stay readable across 6 restrained min-relax frames per condition; E1659A-minus-WT DELTA TOTAL = -2.35 kcal/mol. This supports the sign as a robustness smoke, but production explicit-solvent holo MD / MM-PBSA remains pending. → STRC h01 Phase 9b Min-Relax Multi-Frame MM-GBSA Smoke 2026-04-27

Ranking delta

  • No change. Skeleton plus lite smoke plus tooling/input compatibility plus min-relax multi-frame smoke support the E1659A-favouring sign, but production explicit-solvent holo MD / MM-PBSA has not yet run.

Connections

Graph View

Backlinks