This Title All WIREs
How to cite this WIREs title:
WIREs Comput Mol Sci
Impact Factor: 25.113

Learning molecular potentials with neural networks

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract The potential energy of molecular species and their conformers can be computed with a wide range of computational chemistry methods, from molecular mechanics to ab initio quantum chemistry. However, the proper choice of the computational approach based on computational cost and reliability of calculated energies is a dilemma, especially for large molecules. This dilemma is proved to be even more problematic for studies that require hundreds and thousands of calculations, such as drug discovery. On the other hand, driven by their pattern recognition capabilities, neural networks started to gain popularity in the computational chemistry community. During the last decade, many neural network potentials have been developed to predict a variety of chemical information of different systems. Neural network potentials are proved to predict chemical properties with accuracy comparable to quantum mechanical approaches but with the cost approaching molecular mechanics calculations. As a result, the development of more reliable, transferable, and extensible neural network potentials became an attractive field of study for researchers. In this review, we outlined an overview of the status of current neural network potentials and strategies to improve their accuracy. We provide recent examples of studies that prove the applicability of these potentials. We also discuss the capabilities and shortcomings of the current models and the challenges and future aspects of their development and applications. It is expected that this review would provide guidance for the development of neural network potentials and the exploitation of their applicability. This article is categorized under: Data Science > Artificial Intelligence/Machine Learning Molecular and Statistical Mechanics > Molecular Interactions Software > Molecular Modeling
A typical workflow for fitting neural network potentials
[ Normal View | Magnified View ]
ANI‐2x/MM free energy calculations provide more reliable binding free energy for Tyk2 inhibitor relative to MM calculations reported by Chodera et al.143 (a) absolute binding free energies obtained with ANI‐2x/MM. (b) Absolute binding free energies obtained with OpenFF 1.0.0 force field used with AMBER14SB and TIP3P water model
[ Normal View | Magnified View ]
Torsional potential energy surface of erlotinib reported by Rowley et al.132 (a) Two‐dimensional representation of erlotinib with the torsions of interest. (b) Conformers of erlotinib obtained with CGenFF (green), and ANI‐1ccx/MM model (red) superposed to the crystal structure (blue, PDB ID: 4HJO). (c) Potential energy surfaces obtained with (i) DLPNO‐CCSD(T)/def2‐TZVP//MP2/def2‐TZVP, (ii) ANI‐1ccx/MM and (iii) the CGenFF calculations
[ Normal View | Magnified View ]
Comparison of one‐dimensional potential surface scan generated from DFT, the ANI‐1 potential, and two popular semi‐empirical methods, DFTB and PM6. Each box corresponds to different types of scans. The atoms used to produce the scan coordinate are labeled in the images of the molecules in every subplot. Each figure also lists the RMSE, in the legend, for each method compared with the DFT potential surface. The figure is from Ref. 35
[ Normal View | Magnified View ]
A typical NNP involves terms that are similar to a typical forcefield involving intermolecular and intramolecular terms. The total energy of the system obtained with an NNP can be correlated with the definition of a cut‐off distance (Rcut) during the training, while this cut‐off distance is closely related to the inclusion of short‐range and long‐range interactions
[ Normal View | Magnified View ]
A schematic representation of a high dimensional neural network potential (HDNNP)
[ Normal View | Magnified View ]

Browse by Topic

Software > Molecular Modeling
Molecular and Statistical Mechanics > Molecular Interactions
Computer and Information Science > Visualization

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts