Large Language Models for Inorganic Synthesis Predictions DOI
Seong-Min Kim, Yousung Jung, Joshua Schrier

et al.

Journal of the American Chemical Society, Journal Year: 2024, Volume and Issue: 146(29), P. 19654 - 19659

Published: July 11, 2024

We evaluate the effectiveness of pretrained and fine-tuned large language models (LLMs) for predicting synthesizability inorganic compounds selection precursors needed to perform synthesis. The predictions LLMs are comparable to─and sometimes better than─recent bespoke machine learning these tasks but require only minimal user expertise, cost, time develop. Therefore, this strategy can serve both as an effective strong baseline future studies various chemical applications a practical tool experimental chemists.

Language: Английский

Artificial intelligence and food flavor: How AI models are shaping the future and revolutionary technologies for flavor food development DOI Creative Commons
Zhiyong Cui, Chong Qi, Tianxing Zhou

et al.

Comprehensive Reviews in Food Science and Food Safety, Journal Year: 2025, Volume and Issue: 24(1)

Published: Jan. 1, 2025

Abstract The food flavor science, traditionally reliant on experimental methods, is now entering a promising era with the help of artificial intelligence (AI). By integrating existing technologies AI, researchers can explore and develop new substances in digital environment, saving time resources. More more research will use AI big data to enhance product flavor, improve quality, meet consumer needs, drive industry toward smarter sustainable future. In this review, we elaborate mechanisms recognition their potential impact nutritional regulation. With increase accumulation development internet information technology, databases ingredient have made great progress. These provide detailed content, molecules, chemical properties various compounds, providing valuable support for rapid evaluation components construction screening technology. popularization fields, field has also ushered opportunities. This review explores role enhancing analysis through high‐throughput omics technologies. algorithms offer pathway scientifically formulations, thereby customized meals. Furthermore, it discusses safety challenges into industry.

Language: Английский

Citations

3
AI in single-atom catalysts: a review of design and applications DOI Open Access

Qijun Yu,

Ninggui Ma,

Chihon Leung

et al.

Journal of Materials Informatics, Journal Year: 2025, Volume and Issue: 5(1)

Published: Feb. 12, 2025

Single-atom catalysts (SACs) have emerged as a research frontier in catalytic materials, distinguished by their unique atom-level dispersion, which significantly enhances activity, selectivity, and stability. SACs demonstrate substantial promise electrocatalysis applications, such fuel cells, CO2 reduction, hydrogen production, due to ability maximize utilization of active sites. However, the development efficient stable involves intricate design screening processes. In this work, artificial intelligence (AI), particularly machine learning (ML) neural networks (NNs), offers powerful tools for accelerating discovery optimization SACs. This review systematically discusses application AI technologies through four key stages: (1) Density functional theory (DFT) ab initio molecular dynamics (AIMD) simulations: DFT AIMD are used investigate mechanisms, with high-throughput applications expanding accessible datasets; (2) Regression models: ML regression models identify features that influence performance, streamlining selection promising materials; (3) NNs: NNs expedite known structural models, facilitating rapid assessment potential; (4) Generative adversarial (GANs): GANs enable prediction novel high-performance tailored specific requirements. work provides comprehensive overview current status insights recommendations future advancements field.

Language: Английский

Citations

2
Machine learning in process systems engineering: Challenges and opportunities DOI Creative Commons
Pródromos Daoutidis, Jay H. Lee, Srinivas Rangarajan

et al.

Computers & Chemical Engineering, Journal Year: 2023, Volume and Issue: 181, P. 108523 - 108523

Published: Nov. 22, 2023

Language: Английский

Citations

37
Guided diffusion for inverse molecular design DOI
Tomer Weiss, Eduardo Mayo Yanes, Sabyasachi Chakraborty

et al.

Nature Computational Science, Journal Year: 2023, Volume and Issue: 3(10), P. 873 - 882

Published: Oct. 5, 2023

Language: Английский

Citations

34
MatGPT: A Vane of Materials Informatics from Past, Present, to Future DOI
Zhilong Wang, An Chen, Kehao Tao

et al.

Advanced Materials, Journal Year: 2023, Volume and Issue: 36(6)

Published: Oct. 10, 2023

Abstract Combining materials science, artificial intelligence (AI), physical chemistry, and other disciplines, informatics is continuously accelerating the vigorous development of new materials. The emergence “GPT (Generative Pre‐trained Transformer) AI” shows that scientific research field has entered era intelligent civilization with “data” as basic factor “algorithm + computing power” core productivity. continuous innovation AI will impact cognitive laws methods, reconstruct knowledge wisdom system. This leads to think more about informatics. Here, a comprehensive discussion models infrastructures provided, advances in discovery design are reviewed. With rise paradigms triggered by “AI for Science”, vane informatics: “MatGPT”, proposed technical path planning from aspects data, descriptors, generative models, pretraining directed collaborative training, experimental robots, well efforts preparations needed develop generation informatics, carried out. Finally, challenges constraints faced discussed, order achieve digital, intelligent, automated construction joint interdisciplinary scientists.

Language: Английский

Citations

30
Data Generation for Machine Learning Interatomic Potentials and Beyond DOI
Maksim Kulichenko, Benjamin Nebgen, Nicholas Lubbers

et al.

Chemical Reviews, Journal Year: 2024, Volume and Issue: 124(24), P. 13681 - 13714

Published: Nov. 21, 2024

The field of data-driven chemistry is undergoing an evolution, driven by innovations in machine learning models for predicting molecular properties and behavior. Recent strides ML-based interatomic potentials have paved the way accurate modeling diverse chemical structural at atomic level. key determinant defining MLIP reliability remains quality training data. A paramount challenge lies constructing sets that capture specific domains vast space. This Review navigates intricate landscape essential components integrity data ensure extensibility transferability resulting models. We delve into details active learning, discussing its various facets implementations. outline different types uncertainty quantification applied to atomistic acquisition correlations between estimated true error. role samplers generating informative structures highlighted. Furthermore, we discuss via modified surrogate potential energy surfaces as innovative approach diversify also provides a list publicly available cover

Language: Английский

Citations

15
Directional multiobjective optimization of metal complexes at the billion-system scale DOI
Hannes Kneiding, Ainara Nova, David Balcells

et al.

Nature Computational Science, Journal Year: 2024, Volume and Issue: 4(4), P. 263 - 273

Published: March 29, 2024

Language: Английский

Citations

14
Geometry-complete diffusion for 3D molecule generation and optimization DOI Creative Commons
Alex Morehead, Jianlin Cheng

Communications Chemistry, Journal Year: 2024, Volume and Issue: 7(1)

Published: July 3, 2024

Abstract Generative deep learning methods have recently been proposed for generating 3D molecules using equivariant graph neural networks (GNNs) within a denoising diffusion framework. However, such are unable to learn important geometric properties of molecules, as they adopt molecule-agnostic and non-geometric GNNs their networks, which notably hinders ability generate valid large molecules. In this work, we address these gaps by introducing the Geometry-Complete Diffusion Model (GCDM) molecule generation, outperforms existing molecular models significant margins across conditional unconditional settings QM9 dataset larger GEOM-Drugs dataset, respectively. Importantly, demonstrate that GCDM’s generative process enables model proportion energetically-stable at scale GEOM-Drugs, whereas previous fail do so with features learn. Additionally, show extensions GCDM can not only effectively design specific protein pockets but be repurposed consistently optimize geometry chemical composition stability property specificity, demonstrating new versatility models. Code data freely available on GitHub .

Language: Английский

Citations

12
Has generative artificial intelligence solved inverse materials design? DOI Creative Commons
Hyunsoo Park, Zhenzhu Li, Aron Walsh

et al.

Matter, Journal Year: 2024, Volume and Issue: 7(7), P. 2355 - 2367

Published: July 1, 2024

The directed design and discovery of compounds with pre-determined properties is a long-standing challenge in materials research. We provide perspective on progress toward achieving this goal using generative models for chemical compositions crystal structures based set powerful statistical techniques drawn from the artificial intelligence community. introduce central concepts underpinning crystalline materials. Coverage provided early implementations inorganic crystals adversarial networks variational autoencoders through to ongoing involving autoregressive diffusion models. influence choice representation architecture discussed, along metrics quantifying quality hypothetical produced. While further developments are required enable realistic predictions richer structure property datasets, already proving be complementary traditional strategies.

Language: Английский

Citations

12
Sample efficient reinforcement learning with active learning for molecular design DOI Creative Commons
Michael Dodds, Jeff Guo, Thomas Löhr

et al.

Chemical Science, Journal Year: 2024, Volume and Issue: 15(11), P. 4146 - 4160

Published: Jan. 1, 2024

Reinforcement learning (RL) is a powerful and flexible paradigm for searching solutions in high-dimensional action spaces. However, bridging the gap between playing computer games with thousands of simulated episodes solving real scientific problems complex involved environments (up to actual laboratory experiments) requires improvements terms sample efficiency make most expensive information. The discovery new drugs major commercial application RL, motivated by very large nature chemical space need perform multiparameter optimization (MPO) across different properties.

Language: Английский

Citations

11