PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments, and Partial Charges.

By Oliver T. Unke, Markus Meuwly. 2019

1,056 citations recorded by the source corpus.

In recent years, machine learning (ML) methods have become increasingly popular in computational chemistry. After being trained on appropriate ab initio reference data, these methods allow for accurately predicting the properties of chemical systems, circumventing the need for explicitly solving the electronic Schrödinger equation. Because of their computational efficiency and scalability to large data sets, deep neural networks (DNNs) are a particularly promising ML algorithm for chemical applications. This work introduces PhysNet, a DNN architecture designed for predicting energies, forces, and dipole moments of chemical systems. PhysNet achieves state-of-the-art performance on the QM9, MD17, and ISO17 benchmarks. Further, two new data sets are generated in order to probe the performance of ML models for describing chemical reactions, long-range interactions, and condensed phase systems. It is shown that explicitly including electrostatics in energy predictions is crucial for a qualitatively correct description of the asymptotic regions of a potential energy surface (PES). PhysNet models trained on a systematically constructed set of small peptide fragments (at most eight heavy atoms) are able to generalize to considerably larger proteins like deca-alanine (Ala10): The optimized geometry of helical Ala10 predicted by PhysNet is virtually identical to ab initio results (RMSD = 0.21 Å). By running unbiased molecular dynamics (MD) simulations of Ala10 on the PhysNet-PES in gas phase, it is found that instead of a helical structure, Ala10 folds into a "wreath-shaped" configuration, which is more stable than the helical form by 0.46 kcal mol-1 according to the reference ab initio calculations.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Molecular Biology, Model Architectures, Computational Chemistry & Drug Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Materials Chemistry

In Molecular Biology

In Model Architectures

In Computational Chemistry & Drug Discovery

Citations in the corpus

Directional Message Passing for Molecular Graphs (2020, methodology): “We use 6 state-of-the-art models for comparison: SchNet (Schütt et al., 2017), PhysNet (results based on the reference implementation) (Unke & Meuwly, 2019), provably powerful graph networks (PPGN, results provided by the original authors) (Maron et al., 2019), MEGNet-simple (without auxiliary…”
E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials (2021, methodology): “The same trend can also be observed for various methods on the aspirin molecule in the MD-17 data set (see Figure 8) where across a series of descriptors and regression models (sGDML, FCHL19, and PhysNet [10, 37, 41]) the learning curves show an approximately similar log-log slope (results obtained ”
Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements (2021, background): “1 [8] E(3) Invariant MLP X SchNet [10] E(3) Invariant GNN X DimeNet++ [11, 12] E(3) Invariant GNN X PhysNet [13] E(3) Invariant GNN X Cormorant [14] SE(3) Equivariant GNN X X SE(3)-Transformer [15] SE(3) Equivariant GNN X X NequIP [16] SE(3) Equivariant GNN X X SpookyNet [17] SE(3) Equivariant GNN X”
Machine-learned potentials for next-generation matter simulations (2021, methodology): “To improve this description 56 , machine learning can be used to predict environment-dependent point charges or molecular dipole/multipole moments 57–59 .”