Qmmm Adaptive

from qmmm import QMMMSystem

# Define QM region (reaction center)
qm_atoms = ["C1", "C2", "N3", "O4"]  # Reactive atoms
qm_method = "MOPAC"  # or ORCA, TeraChem for higher accuracy

# Define MM region (protein + solvent)
mm_forcefield = "AMBER14"

# Create hybrid system
system = QMMMSystem(
    structure="enzyme_complex.pdb",
    qm_atoms=qm_atoms,
    qm_method=qm_method,
    mm_forcefield=mm_forcefield,
    qm_charge=0,
    qm_multiplicity=1,
    buffer_distance=15  # Å from QM center
)

# Equilibrate in stages
system.equilibrate_qm_region(steps=5000)     # Fix MM atoms
system.equilibrate_hybrid(steps=10000)       # Allow coupling
system.equilibrate_full(steps=50000)         # Full system

# Define collective variable (reaction coordinate)
cv = ReactionCoordinate(
    atoms=["C1", "C2"],
    type="distance"  # or angle, dihedral, custom
)

# Run metadynamics
metad = Metadynamics(
    system=qmmm_system,
    cv=cv,
    sigma=0.1,           # Width of Gaussian hills
    height=5.0,          # Energy height (kcal/mol)
    stride=100,          # Add hill every N steps
    temperature=300,     # K
    timestep=1.0         # fs
)

trajectory = metad.run(steps=100000)  # 100 ps with ~1000 hills

# Get free energy profile
free_energy = trajectory.get_free_energy()
barriers = free_energy.analyze_barriers()

# Define windows along reaction coordinate
windows = [(1.5 + i*0.1) for i in range(20)]  # 1.5 Å to 3.5 Å

umbrella = UmbrellaSampling(
    system=qmmm_system,
    cv=cv,
    windows=windows,
    force_constant=100.0,  # kcal/mol/Ų
    timestep=1.0,
    temperature=300
)

# Run each window (can be parallelized)
umbrella.run_windows(
    equilibration=5000,   # 5 ps equilibration per window
    production=20000,     # 20 ps production per window
    save_frequency=100,   # Save every 0.1 ps
    n_parallel=4          # Run 4 windows in parallel
)

# Analyze using WHAM
free_energy = umbrella.analyze_wham()

amd = AcceleratedMD(
    system=qmmm_system,
    boost_type="dual",        # Boost both dihedral + total energy
    e_threshold=10.0,         # Boost below E+10 kcal/mol
    alpha_dihedral=5.0,       # Dihedral boost strength
    alpha_total=10.0,         # Total energy boost strength
    timestep=2.0,
    temperature=300
)

trajectory = amd.run(steps=200000)  # 400 ps with acceleration

# Reweight trajectory to remove boost
free_energy = trajectory.reweight_by_boost_energy()

# From metadynamics trajectory
ts_finder = TransitionStateFinder(
    metad_trajectory=metad_traj,
    free_energy_surface=fe_profile
)

# Find TS structure (highest energy on MEP)
ts_structure = ts_finder.find_ts(
    refine=True,
    opt_method="L-BFGS"
)

# Characterize TS
ts_analysis = ts_structure.analyze()
print(f"TS Energy: {ts_analysis['energy']:.2f} kcal/mol")
print(f"Barrier height: {ts_analysis['barrier_forward']:.2f} kcal/mol")
print(f"Barrier height (reverse): {ts_analysis['barrier_reverse']:.2f} kcal/mol")
print(f"TS geometry verified: {ts_analysis['verified']}")

# Example: Serine protease reaction mechanism
# Step 1: Nucleophilic attack (Ser195 on substrate)
# Step 2: Tetrahedral intermediate formation
# Step 3: Acyl-enzyme formation
# Step 4: Water activation and acyl hydrolysis

steps = [
    {
        "name": "Nucleophilic attack",
        "cv": Distance(["Ser195-OG", "C1-substrate"]),
        "expected_barrier": 12.0  # kcal/mol
    },
    {
        "name": "Proton transfer",
        "cv": Distance(["His57-NE", "Proton"]),
        "expected_barrier": 8.0
    },
    {
        "name": "Acylation",
        "cv": Distance(["Ser195-C", "Peptide-N"]),
        "expected_barrier": 5.0
    },
    {
        "name": "Deacylation",
        "cv": Distance(["Water-O", "Acyl-C"]),
        "expected_barrier": 15.0
    }
]

# Run each step with QM/MM
for step in steps:
    result = qmmm.study_step(step)
    print(f"{step['name']}: {result['barrier']:.1f} kcal/mol")