MACE图形神经网络的可转移性为范围校正的 Δ-机器学习潜力 QM/MM 应用程序
Timothy J Giese1, Jinzhe Zeng2,3, Darrin M York1
1Laboratory for Biomolecular Simulation Research, Institute for Quantitative Biomedicine, and Department of Chemistry and Chemical Biology, Rutgers University, Piscataway 08854, New Jersey, United States.
The journal of physical chemistry. B
|May 26, 2025
概括
我们使用图形神经网络开发了一种新的机器学习潜力,用于更准确的分子模拟. 这种方法改善了反应路径和中间体的预测,与以前的方法相比,显示了增强的可转移性.
科学领域:
- 计算化学是一种计算化学.
- 机器学习在化学中的应用
- 量子力学/分子力学 (QM/MM) 模拟
背景情况:
- QM/MM模拟的准确性对于理解复杂的化学反应至关重要.
- 之前的范围校正的 Δ 机器学习潜力 (ΔMLP) 通过纠正能量和力来提高 QM/MM 精度.
- 深度神经网络在提高模拟准确性方面表现出了很大的前景.
研究的目的:
- 通过结合图形神经网络,特别是MACE架构来扩展 ΔMLP 方法.
- 评估AM1/d + MACE模型对基转化反应的可转移性和准确性.
- 在QM/MM模拟中,将MACE与DeepPot-SE (DP) 架构的性能进行比较.
主要方法:
- 训练AM1/d + MACE模型来复制PBE0/6-31G* QM/MM的能量和力.
- 使用不包括在训练集中的反应测试模型的可转移性.
- 计算自由能量表面以评估反应路径的准确性.
- 改变MACE超参数以研究它们对准确性和性能的影响.
主要成果:
- AM1/d + MACE模型准确地复制了目标自由能量表面,在某些情况下表现优于AM1/d + DP模型.
- 最终状态AM1/d + MACE模型正确预测了稳定的五坐标中间体,即使在培训数据中没有类似结构.
- MACE架构证明了 ΔMLP 模型的可转移性得到改善.
- 当使用GPU加速时,AM1/d + MACE模拟被发现比AM1/d QM/MM慢28%.
结论:
- 在QM/MM模拟中,MACE架构为ΔMLP模型提供了更好的可转移性.
- 图形神经网络代表了开发更准确和可转移的机器学习潜力的有希望的方向.
- 开发的AM1/d + MACE模型为研究复杂化学反应 (如基转化) 提供了更可靠的方法.
相关概念视频
01:25Multimachine Stability
237
Multimachine stability analysis is crucial for understanding the dynamics and stability of power systems with multiple synchronous machines. The objective is to solve the swing equations for a network of M machines connected to an N-bus power system.
In analyzing the system, the nodal equations represent the relationship between bus voltages, machine voltages, and machine currents. The nodal equation is given by:
In analyzing the system, the nodal equations represent the relationship between bus voltages, machine voltages, and machine currents. The nodal equation is given by:
237
01:20Mason's Rule
518
Mason's rule is a powerful tool in control systems and signal processing. It simplifies the calculation of transfer functions from signal-flow graphs. This method leverages various elements, including loop gains, forward-path gains, and non-touching loops, to determine the transfer function efficiently.
Loop gain is determined by identifying and tracing a path from a node back to itself. This involves computing the product of branch gains along the loop. Each loop's gain is crucial for...
Loop gain is determined by identifying and tracing a path from a node back to itself. This involves computing the product of branch gains along the loop. Each loop's gain is crucial for...
518
01:29Mechanistic Models: Compartment Models in Algorithms for Numerical Problem Solving
107
Mechanistic models play a crucial role in algorithms for numerical problem-solving, particularly in nonlinear mixed effects modeling (NMEM). These models aim to minimize specific objective functions by evaluating various parameter estimates, leading to the development of systematic algorithms. In some cases, linearization techniques approximate the model using linear equations.
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...
107