Physics-Inspired Structural Representations for Molecules and Materials

Musil, Félix; Grisafi, Andrea; Bartók, Albert P.; Ortner, Christoph; Csänyi, Gábor; Ceriotti, Michele

doi:10.1021/acs.chemrev.1c00021

Cited by 384 publications

(361 citation statements)

References 333 publications

Supporting

Mentioning

361

Contrasting

Order By: Relevance

“…smoothness, additivity and level of locality. 60,61 SOAP descriptors satisfy all these requirements.…”

Section: Descriptors Of Atomic Environmentsmentioning

confidence: 95%

Classifying soft self-assembled materials via unsupervised machine learning of defects

Gardin¹,

Perego²,

Doni³

et al. 2021

Preprint

View full text Add to dashboard Cite

Unlike molecular crystals, soft self-assembled fibres, micelles, vesicles, etc., exhibit a certain order in the arrangement of their constitutive monomers, but also high structural dynamicity and variability. Defects and disordered local domains that continuously form-and-repair in their structures impart to such materials unique adaptive and dynamical properties, which make them, e.g., capable to communicate with each other. However, objective criteria to compare such complex dynamical features and to classify soft supramolecular materials are non-trivial to attain. Here we show a data-driven workflow allowing us to achieve this goal. Building on unsupervised clustering of Smooth Overlap of Atomic Position (SOAP) data obtained from equilibrium molecular dynamics simulations, we can compare a variety of soft supramolecular assemblies via a robust SOAP metric. This provides us with a data-driven "defectometer" to classify different types of supramolecular materials based on the structural dynamics of the ordered/disordered local molecular environments that statistically emerge within them.

show abstract

“…smoothness, additivity and level of locality. 60,61 SOAP descriptors satisfy all these requirements.…”

Section: Descriptors Of Atomic Environmentsmentioning

confidence: 95%

Classifying soft self-assembled materials via unsupervised machine learning of defects

Gardin¹,

Perego²,

Doni³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Important advances have also been made in the application of ML-based methods throughout the field of protein simulations (Noé et al (2020)). Examples include efficient and more accurate potentials [for example Smith et al (2017)], design of coarsegrained potentials (Wang et al (2019)), and work on general frameworks to represent atoms and molecules based on the underlying physics (Schütt et al, 2018;Musil et al, 2021). In addition, subsequent analysis steps, for example the calculation of kinetic properties, can also benefit from ML-based frameworks (Mardt et al, 2018;Olsson and Noé, 2019).…”

Section: Machine Learning and Energy Landscape Explorationmentioning

confidence: 99%

The Energy Landscape Perspective: Encoding Structure and Function for Biomolecules

Röder

Wales

2022

Front. Mol. Biosci.

View full text Add to dashboard Cite

The energy landscape perspective is outlined with particular reference to biomolecules that perform multiple functions. We associate these multifunctional molecules with multifunnel energy landscapes, illustrated by some selected examples, where understanding the organisation of the landscape has provided new insight into function. Conformational selection and induced fit may provide alternative routes to realisation of multifunctionality, exploiting the possibility of environmental control and distinct binding modes.

show abstract

“…The performance of NN potential depends on the quality of the training dataset, which describes the chemical space of PES [18,39]. The whole training dataset of bulk ICM-102 molecules is constructed using both AIMD and VRMD sampling methods, and the detailed configurations are listed in Table S1.…”

Section: Dataset Explorationmentioning

confidence: 99%

Exploring complex reaction networks using neural network-based molecular dynamics simulation

Chu

Luo

Chen

2022

Preprint

View full text Add to dashboard Cite

Ab initio molecular dynamics (AIMD) is an established method to reveal the reactive dynamics of complex systems. However, the computational cost of AIMD restricts the explorable length and time scales to a great extent. Here, we develop a fundamentally different approach using molecular dynamics simulations powered by a neural network potential to investigate complex reaction networks. This potential is trained via a workflow combining AIMD and interactive molecular dynamics in virtual reality (VRMD) to accelerate the sampling of a rare reactive process. The capability of the methodology is demonstrated by achieving a panoramic visualization of the complex reaction networks for decomposition of a novel high explosive (ICM-102), without any predefined reaction coordinates. The study leads to the discovery of new pathways that would be difficult to uncover employing established methods. These results highlight the power of neural network-based molecular dynamics simulations for exploration of complex reaction mechanisms under extreme conditions at the ab initio level, pushing the limit of theoretical and computational chemistry towards the realism and fidelity of experiments.

show abstract

Physics-Inspired Structural Representations for Molecules and Materials

Cited by 384 publications

References 333 publications

Classifying soft self-assembled materials via unsupervised machine learning of defects

Classifying soft self-assembled materials via unsupervised machine learning of defects

The Energy Landscape Perspective: Encoding Structure and Function for Biomolecules

Exploring complex reaction networks using neural network-based molecular dynamics simulation

Contact Info

Product

Resources

About