DeePMD-GNN:用于外部图形神经网络潜力的DeePMD套件插件
Jinzhe Zeng1, Timothy J Giese1, Duo Zhang2,3,4
1Laboratory for Biomolecular Simulation Research, Institute for Quantitative Biomedicine and Department of Chemistry and Chemical Biology, Rutgers University, Piscataway, New Jersey 08854, United States.
Journal of chemical information and modeling
|March 28, 2025
概括
通过将图形神经网络潜能集成到DeePMD套件中,DeePMD-GNN增强了分子模拟,提高了机器学习潜能 (MLP) 和分子动态 (MD) 的互操作性. 这促进了一致的基准测试和科学发现中的更广泛应用.
科学领域:
- 计算化学和材料科学计算化学和材料科学
- 开发先进的模拟工具的开发.
- 机器学习在科学建模中的应用.
背景情况:
- 机器学习潜力 (MLP) 提供高效准确的原子相互作用预测,影响药物发现,催化和材料设计.
- 当前的MLP软件缺乏互操作性,阻碍了一致的基准测试,并需要与分子动力学 (MD) 软件分开的接口.
研究的目的:
- 介绍DeePMD-GNN,一个扩展DeePMD-kit的插件,以支持外部图形神经网络 (GNN) 潜力.
- 为了使GNN模型 (NequIP,MACE) 在DeePMD-kit.com中无集成.
- 为了促进GNN模型在量子力学/分子力学 (QM/MM) 应用中的使用.
主要方法:
- 为DeePMD-kit框架开发DeePMD-GNN插件的开发.
- 将受欢迎的GNN潜力 (NequIP,MACE) 整合到DeePMD-kit生态系统中.
- 对于QM/MM应用程序的范围校正 ΔMLP 形式主义的实施.
主要成果:
- 在DeePMD-GNN中,DeePMD-GNN成功地将外部GNN潜力集成到DeePMD-kit.
- 该插件支持在QM/MM模拟中的GNN模型.
- 在一致的训练条件下,对NequIP,MACE和DPA-2模型进行了基准计算.
结论:
- DeePMD-GNN增强了分子模拟中的机器学习潜力的互操作性.
- 该插件为GNN模型和QM/MM应用程序提供了一个统一的框架.
- 这项工作促进了更一致的基准测试和在科学研究中更广泛地应用先进的MLP.
相关概念视频
01:15Magnetic Vector Potential
In electrostatics, the electric field can be written as the negative gradient of the potential. In magnetostatics, the zero divergence of the magnetic field ensures that the magnetic field can be expressed as the curl of a vector potential. This potential is known as the magnetic vector potential.
Consider an ideal solenoid with n turns per unit length and radius R. If I is the current through the solenoid, the magnetic field inside the solenoid is expressed as the product of vacuum...
Consider an ideal solenoid with n turns per unit length and radius R. If I is the current through the solenoid, the magnetic field inside the solenoid is expressed as the product of vacuum...
01:25Neural Circuits
Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...