ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost
By Justin S. Smith, Olexandr Isayev, Adrian E. Roitberg. 2016
1,346 citations recorded by the source corpus.
We demonstrate how a deep neural network (NN) trained on a data set of quantum mechanical (QM) DFT calculated energies can learn an accurate and transferable atomistic potential for organic molecules containing H, C, N, and O atoms.
Subfields: Atomic and Molecular Physics, and Optics, Computational Theory and Mathematics, Materials Chemistry, Computational Chemistry & Drug Discovery, AI for Physical Sciences
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Atomic and Molecular Physics, and Optics
- Risi Kondor
- Kyle Swanson
- Brandon M. Anderson
- Truong Son Hy
- Evgeny V. Podryabinkin
- Alexander V. Shapeev
- Linfeng Zhang
- Han Wang
In Computational Theory and Mathematics
- Charles Siegel
- Jian Tang
- Abhinav Vishnu
- Garrett B. Goh
- Nathan Oken Hodas
- Kyle Swanson
- Jure Leskovec
- Gisbert Schneider
In Materials Chemistry
- Jian Tang
- Tommi S. Jaakkola
- Jure Leskovec
- Charles Siegel
- Tess E. Smidt
- Nathan Oken Hodas
- Abhinav Vishnu
- Garrett B. Goh
In Computational Chemistry & Drug Discovery
- Tess E. Smidt
- Kyle Swanson
- Sam Schoenholz
- Mario Geiger
- Abhinav Vishnu
- Garrett B. Goh
- Nathan Oken Hodas
- Frank No'e
In AI for Physical Sciences
- Kyle Swanson
- Brandon M. Anderson
- Truong Son Hy
- Sam Schoenholz
- Frank No'e
- Mario Geiger
- Risi Kondor
- Ron O. Dror
Citations in the corpus
- Relevance of Rotationally Equivariant Convolutions for Predicting Molecular Properties (2020, background): “The successes have been characterized by designing bespoke neural networks which have relevant properties “baked-in,” such as parameter sharing across calculations on individual atoms, continuous convolutions, invariance to atomic indexing, and invariance to rotation and translation [4, 5].”
- Cormorant: Covariant Molecular Neural Networks (2019, methodology): “…in using ML methods not just for learning force fields, but also for predicting many other physical/chemical properties of atomic systems across different branches of materials science, chemistry and pharmacology [Montavon et al., 2013, Gilmer et al., 2017, Smith et al., 2017, Yao et al., 2018].”
- Tensor Field Networks: Rotation- and Translation-Equivariant Neural Networks for 3D Point Clouds (2018, background): “Smith et al. [15] additionally uses angular information but does not have general equivariance.”
- Machine learning unifies the modeling of materials and molecules (2017, neutral): “Although there have been several hints that this approach can achieve the accuracy of DFT at a fraction of the cost ( 5 – 11 ), little effort has been put into recovering the generality of quantum mechanics: atomic and molecular descriptors, as well as learning strategies have been optimized for…”
- Learning Neural Generative Dynamics for Molecular Conformation Generation (2021, background): “There have been results showing deep learning speeding up molecular dynamics simulation by learning efficient alternatives to quantum mechanics-based energy calculations (Sch¨utt et al., 2017; Smith et al., 2017).”
- Deep reinforcement learning for de novo drug design (2017, background): “Deep Learning (DL) technologies are beginning to find applications in drug discovery (6–8) including areas of molecular docking (9), transcriptomics (10) , reaction mechanism elucidation (11) , and molecular energy prediction (12, 13).”
- Active Learning of Uniformly Accurate Inter-atomic Potentials for Materials Simulation (2018, background): “Recent progress with machine learning (ML) methods is changing the outlook [7–19].”
- Chemception: A Deep Neural Network with Minimal Chemistry Knowledge Matches the Performance of Expert-developed QSAR/QSPR Models (2017, methodology): “These human-expert features are then used by machine learning algorithms for QSAR and QSPR modeling, which in recent years have also included deep neural networks.(9-23) However, we observed that even in the most recent chemistry literature, deep learning in chemistry still falls under this traditio”
- Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements (2021, methodology): “The structures of the molecules are generated by geometrical optimization, normal mode sampling (NMS) [2], or molecular dynamics.”
- Scalable molecular dynamics on CPU and GPU architectures with NAMD. (2020, background): “…such as polarizable force fields aimed at capturing induction phenomena more faithfully, as well as machine-learning protocols such as the accurate neural network engine for molecular energies (ANAKIN-ME or ANI for short), 246 is a critical advance in order to provide reliable quantitative scales.”
- Deep convolutional networks for quality assessment of protein folds (2018, background): “For small molecules, [22], [23] and [24] have recently developed deep neural networks to predict the molecular energy of a variety of chemical compounds in various conformations (or even various isomeric states).”