Improving molecular force fields across configurational space by combining supervised and unsupervised machine learning.

@article{Fonseca2021ImprovingMF,
  title={Improving molecular force fields across configurational space by combining supervised and unsupervised machine learning.},
  author={Gregory Fonseca and Igor Poltavsky and Valentin Vassilev-Galindo and Alexandre Tkatchenko},
  journal={The Journal of chemical physics},
  year={2021},
  volume={154 12},
  pages={
          124102
        }
}
The training set of atomic configurations is key to the performance of any Machine Learning Force Field (MLFF) and, as such, the training set selection determines the applicability of the MLFF model for predictive molecular simulations. However, most atomistic reference datasets are inhomogeneously distributed across configurational space (CS), and thus, choosing the training set randomly or according to the probability distribution of the data leads to models whose accuracy is mainly defined… 

Figures and Tables from this paper

Opportunities and Challenges for In Silico Drug Discovery at Delta Opioid Receptors

TLDR
Overall, it is posed that there are multiple opportunities to enable in silico drug discovery at the delta opioid receptor to identify novel delta opioid modulators potentially with unique pharmacological properties, such as biased signaling.

Sports Performance Prediction Based on Chaos Theory and Machine Learning

  • Wei Sun
  • Education, Computer Science
    Wireless Communications and Mobile Computing
  • 2022
TLDR
The comparison between shot put data prediction and real data shows that the prediction results given by the model have little deviation and can provide technical services for performance prediction for special sports training.

Simulations of molecular photodynamics in long timescales

Nonadiabatic dynamics simulations in the long timescale (much longer than 10 ps) are the next challenge in computational photochemistry. This paper delimits the scope of what we expect from methods

Linear Atomic Cluster Expansion Force Fields for Organic Molecules: Beyond RMSE

TLDR
It is demonstrated that fast and accurate linear force fields can be built for molecules using the atomic cluster expansion (ACE) framework, and the four- or five-body ACE force fields improve on the accuracy of the empirical force fields by up to a factor of 10, reaching the accuracy typical of recently proposed machine-learning-based approaches.

Choosing the right molecular machine learning potential

TLDR
This work evaluates the performance of popular machine learning potentials in terms of accuracy and computational cost, and delivers structured information for non-specialists in machine learning to guide them through the maze of acronyms, recognize each potential's main features, and judge what they could expect from each one.

Physics-based, neural network force fields for reactive molecular dynamics: Investigation of carbene formation from [EMIM+][OAc-].

TLDR
A general approach following the empirical valence bond framework for constructing ab initio reactive force fields for condensed phase simulations by combining physics-based methods with neural networks (PB/NNs), which ensures the correct asymptotic behavior of electrostatic, polarization, and dispersion interactions.

Training Algorithm Matters for the Performance of Neural Network Potential

TLDR
This article compares the performance of two popular training algorithms, the adaptive moment estimation algorithm and the extended Kalman filter algorithm, using the Behler-Parrinello neural network and two publicly accessible datasets of liquid water and finds that NNPs trained with EKF are more transferable and less sensitive to the value of the learning rate.

Machine Learning Force Fields: Recent Advances and Remaining Challenges.

TLDR
The general aspects of ML techniques in the context of creating ML force fields are discussed, and common features of ML modeling and quantum-mechanical approximations, so-called global and local ML models, and the physical differences behind these two classes of approaches are described.

References

SHOWING 1-10 OF 55 REFERENCES

Machine Learning Force Fields

TLDR
An overview of applications of ML-FFs and the chemical insights that can be obtained from them is given, and a step-by-step guide for constructing and testing them from scratch is given.

On the role of gradients for machine learning of molecular energies and forces

TLDR
This work investigates the convergence of prediction errors of quantum machine learning models for organic molecules trained on energy and force labels, two common data types in molecular simulations and finds that the inclusion of atomic forces in the training data increases the accuracy of the predicted energies and forces 7-fold, compared to models training on energy only.

Deep learning for UV absorption spectra with SchNarc: First steps toward transferability in chemical compound space.

TLDR
The results indicate that transferability is possible for the excited states of the excited ML models, and the previously reported SchNarc approach for excited-state ML is employed and extended.

Quantum Machine Learning with Response Operators in Chemical Compound Space

TLDR
The recently introduced FCHL18 representation is reviewed, in combination with kernel-based QML models to account for response properties by including the corresponding operators in the regression (Christensen et al., J Chem Phys 150(6):064105, 2019), and the use of the electric field differential operator within a Kernel-based operator QML (OQML) approach is discussed.

Automatically growing global reactive neural network potential energy surfaces: A trajectory-free active learning strategy.

TLDR
Through refitting the available analytical PESs of H3 and OH3 reactive systems, an efficient and trajectory-free active learning method is proposed to automatically sample data points for constructing globally accurate reactive potential energy surfaces using neural networks (NNs).

Machine learning for molecular simulation

TLDR
Recent ML methods for molecular simulation are reviewed, with particular focus on (deep) neural networks for the prediction of quantum-mechanical energies and forces, on coarse-grained molecular dynamics, on the extraction of free energy surfaces and kinetics, and on generative network approaches to sample molecular equilibrium structures and compute thermodynamics.

Atomistic Simulation Tools to Study Protein Self-Aggregation.

TLDR
This chapter demonstrates the use of all-atom molecular dynamics simulations to model the self-aggregation of a six-residue amyloidogenic peptide derived from amyloids β, a 39-43 residue-long peptide associated with the pathogenesis of Alzheimer's disease.

Estimation of Protein-Ligand Unbinding Kinetics Using Non-Equilibrium Targeted Molecular Dynamics Simulations

TLDR
Correlating simulations with experimental data from SPR kinetics measurements and X-ray crystallography on two small molecule compound libraries bound to the N-terminal domain of the chaperone Hsp90, it is shown that the mean non-equilibrium work computed in an ensemble of trajectories of enforced ligand unbinding is a promising predictor for lig and unbinding rates.

Atomistic Simulation: A Unique and Powerful Computational Tool for Corrosion Inhibition Research

It is difficult to understand the atomistic information on the interaction at the metal/corrosion inhibitor interface experimentally which is a key to understanding the mechanism by which inhibitors
...