OpenMM 8 安装与上手指南

OpenMM 8简介

OpenMM 支持多种主流力场（如 AMBER、CHARMM、OPLS-AA 等），并可通过 Python 脚本方便地定义体系、设定模拟参数、执行能量最小化与分子动力学计算。它同时提供可定制的积分器（Integrator）与力场插件（Force Plugins），允许研究人员轻松实现新算法或自定义势能函数。

OpenMM 8 在前一版本的基础上进行了显著优化：

• 增强了 CUDA、OpenCL 与 ROCm 的 GPU 后端性能；

• 引入了 新的分子约束与虚拟位点算法，提升了大型体系的稳定性与能量守恒；

• 改进了 序列化（serialization）与插件加载机制，简化了跨平台部署；

• 提供了与 OpenMMTools、MDTraj、ParmEd、PyTorch 等生态系统的更紧密集成，使其能无缝结合深度学习或自由能计算框架。

此外，OpenMM 可直接读取 PDB、AMBER、GROMACS 等常用分子拓扑文件，并可输出标准的 DCD、PDB、XTC 等轨迹格式，方便后续可视化与分析。结合 Jupyter Notebook 或 Python 环境，研究人员能够快速构建、运行并分析从纳米到微秒级的分子动力学模拟。

凭借其开源、跨平台、高性能的特性，OpenMM 已成为分子模拟领域中广受欢迎的核心计算引擎之一，既适用于基础教学与研究，也能支撑高水平的并行化计算与深度学习驱动的分子建模研究。

OpenMM 8 安装

• Python 3.12
• CUDA 12
• OpenMM 8.2.0

安装

conda create -n openmm8.2
conda activate openmm8.2
conda install -c conda-forge pdbfixer
conda install -c conda-forge openmm cuda-version=12

测试安装

python -m openmm.testInstallation

OpenMM 8 上手指南

from openmm.app import *
from openmm import *
from openmm.unit import *
import mdtraj as md
import numpy as np
import matplotlib.pyplot as plt
import sys
import os
import pandas as pd

# Load the PDB structure file
pdb = PDBFile('data/1bna.pdb')

# Define the force field (AMBER14 for protein + TIP3P for water)
forcefield = ForceField('amber14-all.xml', 'amber14/tip3p.xml')

# Create a modeller and add hydrogens and solvent box
modeller = Modeller(pdb.topology, pdb.positions)
modeller.addHydrogens(forcefield)
modeller.addSolvent(forcefield, model='tip3p', 
                    boxSize=Vec3(8, 8, 8)*nanometers, 
                    ionicStrength=0.1*molar)

# Create the system using Particle Mesh Ewald (PME) for electrostatics,
# constrain all hydrogen bonds, and apply a 1.0 nm cutoff
system = forcefield.createSystem(modeller.topology, 
                                 nonbondedMethod=PME, 
                                 constraints=HBonds, 
                                 nonbondedCutoff=1.0*nanometer)

# Define thermodynamic parameters and integrator
temperature = 300*kelvin
pressure = 1*atmosphere
timestep = 2*femtoseconds
integrator = LangevinIntegrator(temperature, 1/picosecond, timestep)

# Add a barostat to maintain constant pressure (NPT ensemble)
barostat = MonteCarloBarostat(pressure, temperature)
system.addForce(barostat)

# Create the simulation object
simulation = Simulation(modeller.topology, system, integrator)
simulation.context.setPositions(modeller.positions)

# --- Energy minimization ---
print("Minimizing energy...")
simulation.minimizeEnergy()

# --- Equilibration phase ---
print("Equilibrating the system...")
simulation.context.setVelocitiesToTemperature(temperature)
simulation.step(1000)  # 1000 steps for equilibration

# --- Production run ---
print("Running production simulation...")
simulation.reporters.append(StateDataReporter('data/output.log', 1000, 
                                              step=True, potentialEnergy=True,
                                              temperature=True, density=True))
simulation.reporters.append(PDBReporter('data/trajectory.pdb', 1000))
simulation.step(5000)  # Production run: 5000 steps (adjust as needed)

# --- Analyze potential energy ---
file_path = 'data/output.log'
column_names = ["Step", "Potential Energy (kJ/mole)", 
                "Temperature (K)", "Density (g/mL)"]

# Load the log file as a CSV
data = pd.read_csv(file_path, skiprows=1, names=column_names)

# Extract Step and Potential Energy columns
steps = data["Step"]
potential_energy = data["Potential Energy (kJ/mole)"]

# Plot potential energy vs step
plt.figure(figsize=(10, 6))
plt.plot(steps, potential_energy, marker='o', linestyle='-', color='b', 
         label='Potential Energy')
plt.title('Potential Energy vs. Step', fontsize=16)
plt.xlabel('Step', fontsize=14)
plt.ylabel('Potential Energy (kJ/mole)', fontsize=14)
plt.grid(True, linestyle='--', alpha=0.6)
plt.legend(fontsize=12)
plt.tight_layout()

# Show the plot
plt.show()

参考资料

• https://openmm.org/
• https://github.com/openmm/openmm
• https://anaconda.org/conda-forge/pdbfixer
• https://anaconda.org/conda-forge/openmm
• Eastman, Peter, Raimondas Galvelis, Raúl P. Peláez, Charlles RA Abreu, Stephen E. Farr, Emilio Gallicchio, Anton Gorenko et al. "OpenMM 8: molecular dynamics simulation with machine learning potentials." The Journal of Physical Chemistry B 128, no. 1 (2023): 109-116.