Efficient hybrid evolutionary optimization of interatomic potential models

Strategies for the construction of machine-learning potentials for accurate and efficient atomic-scale simulations

Recent advances in machine-learning interatomic potentials have enabled the efficient modeling of complex atomistic systems with an accuracy that is comparable to that of conventional quantum-mechanics based methods. At the same time, the construction of new machine-learning potentials can seem a daunting task, as it involves data-science techniques that are not yet common in chemistry and materials science. Here, we provide a tutorial-style overview of strategies and best practices for the construction of artificial neural network (ANN) potentials. We illustrate the most important aspects of (a) data collection, (b) model selection, (c) training and validation, and (d) testing and refinement of ANN potentials on the basis of practical examples. Current research in the areas of active learning and delta learning are also discussed in the context of ANN potentials. This tutorial review aims at equipping computational chemists and materials scientists with the required background knowledge for ANN potential construction and application, with the intention to accelerate the adoption of the method, so that it can facilitate exciting research that would otherwise be challenging with conventional strategies.

Machine learning for interatomic potential models

The use of supervised machine learning to develop fast and accurate interatomic potential models is transforming molecular and materials research by greatly accelerating atomic-scale simulations with little loss of accuracy. Three years ago, Jörg Behler published a perspective in this journal providing an overview of some of the leading methods in this field. In this perspective, we provide an updated discussion of recent developments, emerging trends, and promising areas for future research in this field. We include in this discussion an overview of three emerging approaches to developing machine-learned interatomic potential models that have not been extensively discussed in existing reviews: moment tensor potentials, message-passing networks, and symbolic regression.

Protein folding prediction in the HP model using ions motion optimization with a greedy algorithm

Background

The function of a protein is determined by its native protein structure. Among many protein prediction methods, the Hydrophobic-Polar (HP) model, an ab initio method, simplifies the protein folding prediction process in order to reduce the prediction complexity.

Results

In this study, the ions motion optimization (IMO) algorithm was combined with the greedy algorithm (namely IMOG) and implemented to the HP model for the protein folding prediction based on the 2D-triangular-lattice model. Prediction results showed that the integration method IMOG provided a better prediction efficiency in a HP model. Compared to others, our proposed method turned out as superior in its prediction ability and resilience for most of the test sequences. The efficiency of the proposed method was verified by the prediction results. The global search capability and the ability to escape from the local best solution of IMO combined with a local search (greedy algorithm) to the new algorithm IMOG greatly improve the search for the best solution with reliable protein folding prediction.

Conclusion

Overall, the HP model integrated with IMO and a greedy algorithm as IMOG provides an improved way of protein structure prediction of high stability, high efficiency, and outstanding performance.

Identifying models of dielectric breakdown strength from high-throughput data via genetic programming

The identification of models capable of rapidly predicting material properties enables rapid screening of large numbers of materials and facilitates the design of new materials. One of the leading challenges for computational researchers is determining the best ways to analyze large material data sets to identify models that can rapidly predict a given property. In this paper, we demonstrate the use of genetic programming to generate simple models of dielectric breakdown based on 82 representative dielectric materials. We identified the band gap E_g and phonon cut-off frequency ω_max as the two most relevant features, and new classes of models featuring functions of E_g and ω_max were uncovered. The genetic programming approach was found to outperform other approaches for generating models, and we discuss some of the advantages of this approach.

Evolutionary Algorithm in the Optimization of a Coarse-Grained Force Field

Simulations using residue-scale coarse-grained models of biomolecules are less computationally demanding than simulations employing full-atomistic force fields. However, the coarse-grained models are often difficult and tedious to parametrize for certain applications. Therefore, a systematic and objective method to help develop or adapt the coarse-grained models is needed. We present an automatic method that implements an evolutionary algorithm to find a set of optimal force field parameters for a one-bead coarse-grained model. In addition to an optimized force field, parameter correlations and significance of the potential energy terms can be determined. The method is applied to two classes of problems: the dynamics of an RNA helix and the RNA structure prediction.