Transferability of MACE Graph Neural Network for Range Corrected Δ-Machine Learning Potential QM/MM Applications DOI
Timothy J. Giese, Jinzhe Zeng, Darrin M. York

и другие.

The Journal of Physical Chemistry B, Год журнала: 2025, Номер unknown

Опубликована: Май 26, 2025

We previously introduced a "range corrected" Δ-machine learning potential (ΔMLP) that used deep neural networks to improve the accuracy of combined quantum mechanical/molecular mechanical (QM/MM) simulations by correcting both internal QM and QM/MM interaction energies forces [J. Chem. Theory Comput. 2021, 17, 6993-7009]. The present work extends this approach include graph networks. Specifically, is applied MACE message passing network architecture, series AM1/d + models are trained reproduce PBE0/6-31G* model phosphoryl transesterification reactions. Several designed test transferability varying amount training data calculating free energy surfaces reactions were not included in parameter refinement. compared DP use DeepPot-SE (DP) architecture. found target even instances where exhibit inaccuracies. train "end-state" only from reactant product states 6 Unlike uncorrected profiles, method correctly reproduces stable pentacoordinated phosphorus intermediate though did structures with similar bonding pattern. Furthermore, mechanism hyperparameters defining varied explore their effect on model's performance. 28% slower than when ΔMLP correction performed graphics processing unit. Our results suggest architecture may lead improved transferability.

Язык: Английский

DeePMD-kit v3: A Multiple-Backend Framework for Machine Learning Potentials DOI
Jinzhe Zeng, Duo Zhang, Anyang Peng

и другие.

Journal of Chemical Theory and Computation, Год журнала: 2025, Номер unknown

Опубликована: Май 2, 2025

In recent years, machine learning potentials (MLPs) have become indispensable tools in physics, chemistry, and materials science, driving the development of software packages for molecular dynamics (MD) simulations related applications. These packages, typically built on specific frameworks, such as TensorFlow, PyTorch, or JAX, face integration challenges when advanced applications demand communication across different frameworks. The previous TensorFlow-based implementation DeePMD-kit exemplified these limitations. this work, we introduce version 3, a significant update featuring multibackend framework that supports PaddlePaddle backends, demonstrate versatility architecture through other MLP differentiable force fields. This allows seamless back-end switching with minimal modifications, enabling users developers to integrate using innovation facilitates more complex interoperable workflows, paving way broader MLPs scientific research.

Язык: Английский

Процитировано

1
FE-ToolKit: A Versatile Software Suite for Analysis of High-Dimensional Free Energy Surfaces and Alchemical Free Energy Networks DOI
Timothy J. Giese, Ryan Snyder, Zeke A. Piskulich

и другие.

Journal of Chemical Information and Modeling, Год журнала: 2025, Номер unknown

Опубликована: Май 20, 2025

Free energy simulations play a pivotal role in diverse biological applications, including enzyme design, drug discovery, and biomolecular engineering. The characterization of high-dimensional free surfaces underlying complex enzymatic mechanisms necessitates extensive sampling through umbrella or string method simulations. Accurate ranking target-binding energies across large ligand libraries relies on comprehensive alchemical calculations organized into thermodynamic networks. predictive accuracy these methods hinges robust, scalable tools for networkwide data analysis extraction physical properties from heterogeneous simulation data. Here, we introduce FE-ToolKit, versatile software suite the automated surfaces, minimum paths, networks (thermodynamic graphs).

Язык: Английский

Процитировано

0
Transferability of MACE Graph Neural Network for Range Corrected Δ-Machine Learning Potential QM/MM Applications DOI
Timothy J. Giese, Jinzhe Zeng, Darrin M. York

и другие.

The Journal of Physical Chemistry B, Год журнала: 2025, Номер unknown

Опубликована: Май 26, 2025

We previously introduced a "range corrected" Δ-machine learning potential (ΔMLP) that used deep neural networks to improve the accuracy of combined quantum mechanical/molecular mechanical (QM/MM) simulations by correcting both internal QM and QM/MM interaction energies forces [J. Chem. Theory Comput. 2021, 17, 6993-7009]. The present work extends this approach include graph networks. Specifically, is applied MACE message passing network architecture, series AM1/d + models are trained reproduce PBE0/6-31G* model phosphoryl transesterification reactions. Several designed test transferability varying amount training data calculating free energy surfaces reactions were not included in parameter refinement. compared DP use DeepPot-SE (DP) architecture. found target even instances where exhibit inaccuracies. train "end-state" only from reactant product states 6 Unlike uncorrected profiles, method correctly reproduces stable pentacoordinated phosphorus intermediate though did structures with similar bonding pattern. Furthermore, mechanism hyperparameters defining varied explore their effect on model's performance. 28% slower than when ΔMLP correction performed graphics processing unit. Our results suggest architecture may lead improved transferability.

Язык: Английский

Процитировано

0