Contact map based crystal structure prediction using global optimization

Hu, Jianjun; Yang, Wenhui; Dong, Rongzhi; Li, Yuxin; Li, Xiang; Li, Shaobo; Siriwardane, Edirisuriya M. Dilanga

doi:10.1039/d0ce01714k

Cited by 20 publications

(30 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method is conceptually similar to our decoder, but it requires additional force data and can only be applied to 1-2 elements due to the exponentially increased data need for more elements. Remotely related works include generating contact maps from chemical compositions Yang et al, 2021) and building generative models only for compositions Pathak et al, 2020;Dan et al, 2020).…”

Section: Related Workmentioning

confidence: 99%

Crystal Diffusion Variational Autoencoder for Periodic Material Generation

Xie¹,

Fu²,

Ganea³

et al. 2021

Preprint

View full text Add to dashboard Cite

Generating the periodic structure of stable materials is a long-standing challenge for the material design community. This task is difficult because stable materials only exist in a low-dimensional subspace of all possible periodic arrangements of atoms: 1) the coordinates must lie in the local energy minimum defined by quantum mechanics, and 2) global stability also requires the structure to follow the complex, yet specific bonding preferences between different atom types. Existing methods fail to incorporate these factors and often lack proper invariances. We propose a Crystal Diffusion Variational Autoencoder (CDVAE) that captures the physical inductive bias of material stability. By learning from the data distribution of stable materials, the decoder generates materials in a diffusion process that moves atomic coordinates towards a lower energy state and updates atom types to satisfy bonding preferences between neighbors. Our model also explicitly encodes interactions across periodic boundaries and respects permutation, translation, rotation, and periodic invariances. We significantly outperform past methods in three tasks: 1) reconstructing the input structure, 2) generating valid, diverse, and realistic materials, and 3) generating materials that optimize a specific property. We also provide several standard datasets and evaluation metrics for the broader machine learning community.

show abstract

Section: Related Workmentioning

confidence: 99%

Crystal Diffusion Variational Autoencoder for Periodic Material Generation

Xie¹,

Fu²,

Ganea³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In our previous work [4], the crystal structure prediction problem can be mapped to two related problems: 1) prediction of the contact map of atoms; 2) the atomic coordinate reconstruction from the contact map using global optimization algorithms. We have applied both genetic algorithms and differential evolution algorithms [37] for the coordinate reconstruction.…”

Section: B Age-fitness Based Multi-objective Genetic Algorithm For Cr...mentioning

confidence: 99%

“…The discovery and development of new materials are fundamental to the progress of technology. There are several promising approaches for exploring new materials including crystal structure predictions [1][2][3][4], generative machine learning models [5][6][7][8], inverse materials design [5,9], and first-principles [10][11][12] calculation based structural tinkering. Materials Genome Initiative attempts to use data-driven methods [13][14][15] to help discover new material science research paradigms and accelerate the design and exploration of new materials.…”

Section: Introductionmentioning

confidence: 99%

“…However, predicting the crystal structure of the given chemical composition is a major difficulty that has merits decades of research [21]. We previously proposed a contact map-based crystal structure prediction method AlphaCrystal [24], which predicts the contact map [25], space group, and lattice constants of the material through deep learning methods, and then uses global optimization algorithms such as GA and PSO to maximize the match between the contact map of the predicted structure and the predicted contact map by searching for the Wyckoff positions. Our experiments proved that the geometric constraints [26][27][28] in the crystal structure help the crystal structure reconstruction.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Crystal structure prediction using age-fitness multi-objective genetic algorithm and coordination number constraints

Yang,

Siriwardane,

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Crystal structure prediction (CSP) has emerged as one of the most important approaches for discovering new materials. CSP algorithms based on evolutionary algorithms and particle swarm optimization have discovered a great number of new materials. However, these algorithms based on ab initio calculation of free energy are inefficient. Moreover, they have severe limitations in terms of scalability. We recently proposed a promising crystal structure prediction method based on atomic contact maps, using global optimization algorithms to search for the Wyckoff positions by maximizing the match between the contact map of the predicted structure and the contact map of the true crystal structure. However, our previous contact map based CSP algorithms have two major limitations: (1) the loss of search capability due to getting trapped in local optima; (2) it only uses the connection of atoms in the unit cell to predict the crystal structure, ignoring the chemical environment outside the unit cell, which may lead to unreasonable coordination environments. Herein we propose a novel multi-objective genetic algorithms for contact map-based crystal structure prediction by optimizing three objectives, including contact map match accuracy, the individual age, and the coordination number match. Furthermore, we assign the age values to all the individuals of the GA and try to minimize the age aiming to avoid the premature convergence problem. Our experimental results show that compared to our previous CMCrystal algorithm, our multi-objective crystal structure prediction algorithm (CMCrystalMOO) can reconstruct the crystal structure with higher quality and alleviate the problem of premature convergence.

show abstract

“…The latest approach uses deep learning to predict the contact maps or distance matrix, which can then be used to reconstruct the full 3D protein structure with a high accuracy . In our prior work, we have shown that given a crystal’s space group, its lattice constants, and the contact map, all of which can be predicted, we can reconstruct the crystal structure atom coordinates using global optimization algorithms such as genetic algorithms (GAs), particle swarm optimization, or Bayesian optimization. However, it is known that pairwise distances between atoms within a crystal structure usually are conserved across different compounds, so that machine learning models can be built to predict the distance matrix of a given composition.…”

Section: Introductionmentioning

confidence: 99%

Distance Matrix-Based Crystal Structure Prediction Using Evolutionary Algorithms

Yang

Siriwardane

2020

J. Phys. Chem. A

Self Cite

View full text Add to dashboard Cite

Crystal structure prediction (CSP) for inorganic materials is one of the central and most challenging problems in materials science and computational chemistry. This problem can be formulated as a global optimization problem in which global search algorithms such as genetic algorithms (GAs) and particle swarm optimization have been combined with first-principles free-energy calculations to predict crystal structures given only the material composition or a chemical system. These DFT-based ab initio CSP algorithms are computationally demanding and can usually be used only to predict crystal structures of relatively small systems. The vast coordinate space and the expensive DFT free-energy calculations limit their inefficiency and scalability. On the other hand, a similar structure prediction problem has been intensively investigated in parallel in the protein structure prediction (PSP) community of bioinformatics, in which the dominating predictors are knowledge-based approaches including homology modeling and threading that exploit known protein structures. Surprisingly, the CSP field has mainly focused on ab initio approaches in the past decade. Inspired by the knowledge-rich PSP approaches, herein, we explore whether known geometric constraints such as the pairwise atomic distances of a target crystal material can help predict/reconstruct its structure given its space group and lattice information. We propose DMCrystal, a GA-based crystal structure reconstruction algorithm based on predicted pairwise atomic distances. Based on extensive experiments, we show that the predicted distance matrix can dramatically help reconstruct the crystal structure and usually achieves much better performance than that of CMCrystal, an atomic contact map-based CSP algorithm. This implies that the knowledge of atomic interaction information learned from the existing materials can be used to significantly improve the CSP in terms of both speed and quality.

show abstract

Contact map based crystal structure prediction using global optimization

Abstract: Crystal structure prediction is now playing an increasingly important role in the discovery of new materials or crystal engineering.

Cited by 20 publications

References 40 publications

Crystal Diffusion Variational Autoencoder for Periodic Material Generation

Crystal Diffusion Variational Autoencoder for Periodic Material Generation

Crystal structure prediction using age-fitness multi-objective genetic algorithm and coordination number constraints

Distance Matrix-Based Crystal Structure Prediction Using Evolutionary Algorithms

Contact Info

Product

Resources

About