molecular-dynamics

Skill documentation

Molecular Dynamics

Overview

Molecular dynamics (MD) simulation computationally models the time evolution of molecular systems by integrating Newton's equations of motion. This skill covers two complementary tools:

OpenMM (https://openmm.org/): High-performance MD simulation engine with GPU support, Python API, and flexible force field support
MDAnalysis (https://mdanalysis.org/): Python library for reading, writing, and analyzing MD trajectories from all major simulation packages

Reviewed versions: OpenMM 8.6.1 and MDAnalysis 2.10.0. Reference-platform smoke checks use small synthetic systems; GPU performance and scientific convergence are not established by them. OpenMM latest API pages identify a development build, so the installed 8.6.1 API is the executable reference.

Installation into a dedicated environment:

uv venv --python 3.12 .venv-md
uv pip install --python .venv-md/bin/python "openmm==8.6.1" "MDAnalysis==2.10.0" matplotlib pandas
# Windows: use .venv-md/Scripts/python.exe instead

When to Use This Skill

Use molecular dynamics when:

Protein stability analysis: How does a mutation affect protein dynamics?
Drug binding simulations: Characterize binding mode and residence time of a ligand
Conformational sampling: Explore protein flexibility and conformational changes
Protein-protein interaction: Model interface dynamics and binding energetics
RMSD/RMSF analysis: Quantify structural fluctuations from a reference structure
Free energy estimation: Compute binding free energy or conformational free energy
Membrane simulations: Model proteins in lipid bilayers
Intrinsically disordered proteins: Study IDR conformational ensembles

Core Workflow: OpenMM Simulation

Examples require system-specific preparation and validation. Review biological assembly, alternate locations, missing loops, termini, protonation, disulfides, ligands and cofactors before parameterization. An MD trajectory alone does not establish binding free energy or residence time.

The functions below form one Python module: execute their import blocks together. Use a new output prefix for every stage to avoid overwriting earlier results.

1. System Preparation

from openmm.app import *
from openmm import *
from openmm.unit import *

def prepare_system_from_pdb(pdb_file, forcefield_name="amber14-all.xml",
                              water_model="amber14/tip3pfb.xml", ph=7.0):
    """
    Prepare an OpenMM system from a PDB file.

    Args:
        pdb_file: Path to cleaned PDB file (use PDBFixer for raw PDB files)
        forcefield_name: Force field XML file
        water_model: Water model XML file

    Returns:
        modeller, system
    """
    # Load PDB
    pdb = PDBFile(pdb_file)

    # Load force field
    forcefield = ForceField(forcefield_name, water_model)

    # Add hydrogens and solvate
    modeller = Modeller(pdb.topology, pdb.positions)
    modeller.addHydrogens(forcefield, pH=ph)

    # TIP3P geometry also serves TIP3P-FB; XML supplies the parameters.
    # 10 Å padding; added salt excludes neutralizing counterions.
    modeller.addSolvent(
        forcefield,
        model='tip3p',
        padding=10*angstroms,
        ionicStrength=0.15*molar, neutralize=True
    )

    print(f"System: {modeller.topology.getNumAtoms()} atoms, "
          f"{modeller.topology.getNumResidues()} residues")

    # Create system
    system = forcefield.createSystem(
        modeller.topology,
        nonbondedMethod=PME,         # Particle Mesh Ewald for long-range electrostatics
        nonbondedCutoff=1.0*nanometer,
        constraints=HBonds,           # Constrain bonds involving H (not intermolecular H-bonds)
        rigidWater=True,
        ewaldErrorTolerance=0.0005
    )

    return modeller, system

2. Energy Minimization

from openmm.app import *
from openmm import *
from openmm.unit import *

def minimize_energy(modeller, system, output_pdb="minimized.pdb",
                     max_iterations=1000, tolerance=10.0, platform_name=None):
    """
    Energy minimize the system to remove steric clashes.

    Args:
        modeller: Modeller object with topology and positions
        system: OpenMM System
        output_pdb: Path to save minimized structure
        max_iterations: Maximum minimization steps
        tolerance: Convergence criterion in kJ/mol/nm

    Returns:
        simulation object with minimized positions
    """
    # Keep the integrator at 2 fs for the subsequent NVT/NPT examples.
    integrator = LangevinMiddleIntegrator(300*kelvin, 1/picosecond, 0.002*picoseconds)

    # Automatic platform selection, or an explicit tested platform (e.g. CPU).
    # A registered GPU plugin does not prove a working device or driver.
    platform = Platform.getPlatformByName(platform_name) if platform_name else None
    simulation = Simulation(modeller.topology, system, integrator, platform)
    simulation.context.setPositions(modeller.positions)

    # Check initial energy
    state = simulation.context.getState(getEnergy=True)
    print(f"Initial energy: {state.getPotentialEnergy()}")

    # Minimize
    simulation.minimizeEnergy(
        tolerance=tolerance*kilojoules_per_mole/nanometer,
        maxIterations=max_iterations
    )

    state = simulation.context.getState(getEnergy=True, getPositions=True)
    print(f"Minimized energy: {state.getPotentialEnergy()}")

    # Save minimized structure
    with open(output_pdb, 'w') as f:
        PDBFile.writeFile(simulation.topology, state.getPositions(), f)

    return simulation

3. NVT Equilibration

from openmm.app import *
from openmm import *
from openmm.unit import *

def run_nvt_equilibration(simulation, n_steps=50000, temperature=300,
                            report_interval=1000, output_prefix="nvt"):
    """
    NVT equilibration: constant N, V, T.
    Equilibrate velocities to target temperature.

    Args:
        simulation: OpenMM Simulation (after minimization)
        n_steps: Number of MD steps (50000 × 2fs = 100 ps)
        temperature: Temperature in Kelvin
        report_interval: Steps between data reports
        output_prefix: File prefix for trajectory and log
    """
    # This example is unrestrained; add validated restraints before Context creation.
    if any("Barostat" in type(f).__name__ and f.getFrequency() > 0
           for f in simulation.system.getForces()):
        raise ValueError("NVT requires all barostats disabled or absent")

    # Set both the thermostat target and initial velocities.
    simulation.integrator.setTemperature(temperature*kelvin)
    simulation.context.setVelocitiesToTemperature(temperature*kelvin)

    # Add reporters
    simulation.reporters = []

    # Log file
    simulation.reporters.append(
        StateDataReporter(
            f"{output_prefix}_log.txt",
            report_interval,
            step=True,
            potentialEnergy=True,
            kineticEnergy=True,
            temperature=True,
            volume=True,
            speed=True
        )
    )

    # DCD trajectory (compact binary format)
    simulation.reporters.append(
        DCDReporter(f"{output_prefix}_traj.dcd", report_interval)
    )

    duration = (n_steps * simulation.integrator.getStepSize()).value_in_unit(picoseconds)
    print(f"Running NVT equilibration: {n_steps} steps ({duration:.1f} ps)")
    simulation.step(n_steps)
    print("NVT equilibration complete")

    return simulation

4. NPT Equilibration and Production

Call this function first with a dedicated NPT equilibration prefix. Inspect density, energy, structure and replicate stability before a separate production call; the default duration is an example, not an equilibration or convergence criterion.

def run_npt_production(simulation, n_steps=500000, temperature=300, pressure=1.0,
                        report_interval=5000, output_prefix="npt"):
    """
    NPT production run: constant N, P, T.

    Args:
        n_steps: Production steps (500000 × 2fs = 1 ns)
        temperature: Temperature in Kelvin
        pressure: Pressure in bar
        report_interval: Steps between reports
    """
    # Keep the Langevin thermostat and barostat at the same temperature.
    simulation.integrator.setTemperature(temperature*kelvin)
    system = simulation.system
    barostats = [f for f in system.getForces() if "Barostat" in type(f).__name__]
    if not system.usesPeriodicBoundaryConditions():
        raise ValueError("NPT requires a periodic system")
    if not barostats:
        system.addForce(MonteCarloBarostat(pressure*bar, temperature*kelvin, 25))
        simulation.context.reinitialize(preserveState=True)
    elif len(barostats) != 1 or not isinstance(barostats[0], MonteCarloBarostat):
        raise ValueError("This example supports one isotropic MonteCarloBarostat")
    else:
        barostats[0].setDefaultPressure(pressure*bar)
        barostats[0].setDefaultTemperature(temperature*kelvin)
        barostats[0].setFrequency(25)
    # Existing Context parameters must also be updated on repeated calls.
    simulation.context.setParameter(MonteCarloBarostat.Pressure(), pressure)
    simulation.context.setParameter(MonteCarloBarostat.Temperature(), temperature)

    # Update reporters
    simulation.reporters = []
    simulation.reporters.append(
        StateDataReporter(
            f"{output_prefix}_log.txt",
            report_interval,
            step=True,
            potentialEnergy=True,
            temperature=True,
            density=True,
            speed=True
        )
    )
    simulation.reporters.append(
        DCDReporter(f"{output_prefix}_traj.dcd", report_interval)
    )

    # Save checkpoints
    simulation.reporters.append(
        CheckpointReporter(f"{output_prefix}_checkpoint.chk", 50000)
    )

    duration = (n_steps * simulation.integrator.getStepSize()).value_in_unit(nanoseconds)
    print(f"Running NPT stage: {n_steps} steps ({duration:.3f} ns)")
    simulation.step(n_steps)
    simulation.saveCheckpoint(f"{output_prefix}_checkpoint.chk")
    simulation.saveState(f"{output_prefix}_state.xml")
    print("NPT stage complete")
    return simulation

Trajectory Analysis with MDAnalysis

1. Load Trajectory

Use the solvated topology with exactly the DCD atom count and order (e.g. the minimized.pdb above), not the original unsolvated input. MDAnalysis converts lengths to Å and time to ps by default; OpenMM bare coordinates use nm. Preserve frame box vectors and actual timestamps; do not infer time from frame number.

import MDAnalysis as mda
from MDAnalysis.analysis import rms, align, contacts
import numpy as np
import matplotlib.pyplot as plt

def load_trajectory(topology_file, trajectory_file):
    """
    Load an MD trajectory with MDAnalysis.

    Args:
        topology_file: PDB, PSF, or other topology file
        trajectory_file: DCD, XTC, TRR, or other trajectory
    """
    u = mda.Universe(topology_file, trajectory_file)
    print(f"Universe: {u.atoms.n_atoms} atoms, {u.trajectory.n_frames} frames")
    first, last = u.trajectory[0].time, u.trajectory[-1].time
    print(f"Time range: {first:g} to {last:g} ps")
    u.trajectory[0]
    return u

2. RMSD Analysis

def compute_rmsd(u, selection="backbone", reference_frame=0):
    """
    Compute RMSD of selected atoms relative to reference frame.

    Args:
        u: MDAnalysis Universe
        selection: Atom selection string (MDAnalysis syntax)
        reference_frame: Frame index for reference structure

    Returns:
        numpy array with columns [frame index, time in ps, RMSD in Angstroms]
    """
    # RMSD performs its own fit; avoid rotating the stored trajectory here.
    # Make the selected molecule whole before analysis of periodic trajectories.
    R = rms.RMSD(u, select=selection, ref_frame=reference_frame)
    R.run()

    rmsd_data = R.results.rmsd  # columns: frame, time, RMSD
    return rmsd_data

def plot_rmsd(rmsd_data, title="RMSD over time", output_file="rmsd.png"):
    """Plot RMSD over simulation time."""
    fig, ax = plt.subplots(figsize=(10, 4))
    ax.plot(rmsd_data[:, 1] / 1000, rmsd_data[:, 2], 'b-', linewidth=0.5)
    ax.set_xlabel("Time (ns)")
    ax.set_ylabel("RMSD (Å)")
    ax.set_title(title)
    ax.axhline(rmsd_data[:, 2].mean(), color='r', linestyle='--',
               label=f'Mean: {rmsd_data[:, 2].mean():.2f} Å')
    ax.legend()
    plt.tight_layout()
    plt.savefig(output_file, dpi=150)
    return fig

3. RMSF Analysis (Per-Residue Flexibility)

Make molecules whole and align a separate structural-analysis copy before RMSF; RMSF does not align. Do not run periodic contacts on a rotated trajectory whose box was not rotated. See analysis reference for alignment, PCA, DSSP, hydrogen bonds and population-derived free energy.

def compute_rmsf(u, selection="protein and name CA", start_frame=0):
    """
    Compute per-residue RMSF (flexibility).

    Returns:
        residue_keys, rmsf_values; keys are (resindex, segid, resid, resname)
    """
    # Select atoms
    atoms = u.select_atoms(selection)

    if not len(atoms):
        raise ValueError("RMSF selection is empty")
    # A multi-atom selection returns mean atomic RMSF, not COM RMSF.
    R = rms.RMSF(atoms)
    R.run(start=start_frame)

    # Average by residue
    resids = []
    rmsf_per_res = []
    for res in u.select_atoms(selection).residues:
        res_atoms = res.atoms.intersection(atoms)
        if len(res_atoms) > 0:
            resids.append((res.ix, res.segid, res.resid, res.resname))
            # RMSF is indexed within the selected AtomGroup, not the Universe.
            rmsf_per_res.append(R.results.rmsf[atoms.resindices == res.ix].mean())

    return resids, np.array(rmsf_per_res)

4. Protein-Ligand Contacts

def analyze_contacts(u, protein_sel="protein", ligand_sel="resname LIG",
                      radius=4.5, start_frame=0, periodic=True):
    """
    Track protein-ligand contacts over trajectory.

    Args:
        radius: Atom-pair cutoff in Angstroms; any pair defines a residue contact.
        periodic: Require box data and use minimum-image distances.
    Returns sets of (resindex, segid, resid, resname), one per analyzed frame.
    """
    from MDAnalysis.lib.distances import distance_array

    protein = u.select_atoms(protein_sel)
    ligand = u.select_atoms(ligand_sel)

    if not len(protein) or not len(ligand):
        raise ValueError("Protein and ligand selections must both be nonempty")
    if not np.isfinite(radius) or radius <= 0:
        raise ValueError("radius must be finite and positive")
    contact_frames = []
    for ts in u.trajectory[start_frame:]:
        if periodic and (ts.dimensions is None or not np.all(np.isfinite(ts.dimensions))
                         or np.any(ts.dimensions[:3] <= 0)):
            raise ValueError("Periodic contacts require valid frame box dimensions")
        # This includes hydrogen atoms unless the caller selects heavy atoms.
        distances = contacts.contact_matrix(
            distance_array(protein.positions, ligand.positions, box=ts.dimensions if periodic else None),
            radius,
        )
        contact_residues = set()
        for i in range(distances.shape[0]):
            if distances[i].any():
                res = protein[i].residue
                contact_residues.add((res.ix, res.segid, res.resid, res.resname))
        contact_frames.append(contact_residues)

    return contact_frames

Force Fields and Preparation

Choose a validated protein/ligand/water/ion combination for the scientific system. The working example retains AMBER14/TIP3P-FB; it is not a universal recommendation. Current OpenMM also bundles amber19-all.xml (ff19SB, DNA OL21, RNA OL3, lipid21), and charmm36_2024.xml with its own charmm36_2024/water.xml. Generic TIP3P and CHARMM-modified TIP3P are not interchangeable. IDP ensembles are especially sensitive to protein-water balance. Four-site waters need extra particles and a matching Modeller geometry, unlike the three-site example above.

AMBER protein XML does not parameterize arbitrary ligands. Use a reviewed GAFF2 or OpenFF route with explicit stereochemistry, protonation, bond orders, atom mapping and charge method. A ligand-only Interchange is not a protein-ligand system. See preparation reference for conservative PDBFixer and OpenFF examples, installation requirements and current release details.

Validation and Provenance

Start with minimization; finite energy does not establish a valid structure.
Use staged NVT → NPT equilibration → production. This example has no restraints.
A 2 fs step with constrained H-containing bonds is a starting choice. Larger steps/HMR require integrator, stability and observable validation for that system.
Use per-frame boxes for periodic distances. Make molecules whole for structural observables using trusted bond topology; unwrapping cannot infer missing bonds.
Assess burn-in, autocorrelation, effective samples, replicas and uncertainties. Short stable temperature/density traces do not establish conformational sampling.
Save topology/atom order, System/Integrator XML, force-field files/versions, seeds, platform/precision, parameters and analysis selections. Checkpoints are platform/hardware/version specific; XML states are more portable but do not retain all internal random-generator state. Test restarting the actual setup.

Additional Resources

OpenMM documentation: https://openmm.org/documentation.html
MDAnalysis user guide: https://docs.mdanalysis.org/
GROMACS (alternative MD engine): https://manual.gromacs.org/
NAMD (alternative): https://www.ks.uiuc.edu/Research/namd/
CHARMM-GUI (web-based system builder): https://charmm-gui.org/
AmberTools (free Amber tools): https://ambermd.org/AmberTools.php
OpenMM paper: Eastman P et al. (2017) PLOS Computational Biology. PMID: 28278240
MDAnalysis paper: Michaud-Agrawal N et al. (2011) J Computational Chemistry. PMID: 21500218

Use this skill in your project

A prompt to adapt

Skill documentation