Data Generation for Machine Learning Interatomic Potentials and Beyond

Chemical Reviews, Год журнала: 2024, Номер 124(24), С. 13681 - 13714

Опубликована: Ноя. 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

Язык: Английский

---

Directional multiobjective optimization of metal complexes at the billion-system scale

Nature Computational Science, Год журнала: 2024, Номер 4(4), С. 263 - 273

Опубликована: Март 29, 2024

Язык: Английский

---

Zinc metal complexes synthesized by a green method as a new approach to alter the structural and optical characteristics of PVA: new field for polymer composite fabrication with controlled optical band gap

RSC Advances, Год журнала: 2024, Номер 14(36), С. 26362 - 26387

Опубликована: Янв. 1, 2024

The current study employed a novel approach to design polymer composites with modified structural and declined optical band gaps. results obtained in the present work for can be considered an original method make new field research based on green chemistry. Natural dyes extracted from tea were mixed hydrated zinc acetate (Zn(CH

Язык: Английский

---

Molecular representations in bio-cheminformatics

Memetic Computing, Год журнала: 2024, Номер 16(3), С. 519 - 536

Опубликована: Июль 20, 2024

Abstract Molecular representations have essential roles in bio-cheminformatics as they facilitate the growth of machine learning applications numerous sub-domains biology and chemistry, especially drug discovery. These transform structural chemical information molecules into machine-readable formats that can be efficiently processed by computer programs. In this paper, we present a comprehensive review, providing readers with diverse perspectives on strengths weaknesses well-known molecular representations, along their respective categories implementation sources. Moreover, provide summary applicability these de novo design, property prediction, reactions. Besides, for macromolecules are discussed highlighted pros cons. By addressing aspects, aim to offer valuable resource significant role advancing its related domains.

Язык: Английский

---

Using Machine Learning to Predict the Antibacterial Activity of Ruthenium Complexes**

Angewandte Chemie International Edition, Год журнала: 2023, Номер 63(10)

Опубликована: Дек. 13, 2023

Rising antimicrobial resistance (AMR) and lack of innovation in the antibiotic pipeline necessitate novel approaches to discovering new drugs. Metal complexes have proven be promising compounds, but number studied compounds is still low compared millions organic molecules investigated so far. Lately, machine learning (ML) has emerged as a valuable tool for guiding design small molecules, potentially even low-data scenarios. For first time, we extend application ML discovery metal-based medicines. Utilising 288 modularly synthesized ruthenium arene Schiff-base their antibacterial properties, series models were trained. The perform well are used predict activity 54 compounds. These displayed 5.7x higher hit-rate (53.7 %) against methicillin-resistant Staphylococcus aureus (MRSA) original library (9.4 %), demonstrating that can applied improve success-rates search metalloantibiotics. This work paves way more ambitious applications field drug discovery.

Язык: Английский

---

Automated Transition Metal Catalysts Discovery and Optimisation with AI and Machine Learning

ChemCatChem, Год журнала: 2024, Номер 16(10)

Опубликована: Янв. 5, 2024

Abstract Significant progress has been made in recent years the use of AI and Machine Learning (ML) for catalyst discovery optimisation. The effectiveness ML data science techniques was demonstrated predicting optimising enantioselectivity regioselectivity catalytic reactions through optimisation ligands, counterions reaction conditions. Direct new catalysts/reactions is more difficult requires efficient exploration transition metal chemical space. A range computational descriptor generation, ranging from molecular mechanics to DFT methods, have successfully demonstrated, often conjunction with reduce cost associated TS calculations. Complex aspects reactions, such as solvent, temperature, etc., also incorporated into workflow.

Язык: Английский

---

A Deep Generative Model for the Inverse Design of Transition Metal Ligands and Complexes

JACS Au, Год журнала: 2025, Номер 5(5), С. 2294 - 2308

Опубликована: Апрель 23, 2025

Deep generative models yielding transition metal complexes (TMCs) remain scarce despite the key role of these compounds in industrial catalytic processes, anticancer therapies, and energy transition. Compared to drug discovery within chemical space organic molecules, TMCs pose further challenges, including encoding bonds higher complexity need optimize multiple properties. In this work, we developed a model for inverse design ligands complexes, based on junction tree variational autoencoder (JT-VAE). After implementing SMILES-based metal-ligand bonds, was trained with tmQMg-L ligand library, allowing generation thousands novel, highly diverse monodentate (κ1) bidentate (κ2) ligands, imines, phosphines, carbenes. Further, generated were labeled two target properties reflecting stability electron density associated homoleptic iridium TMCs: HOMO-LUMO gap (ϵ) charge center (q Ir). This data used implement conditional that from prompt, single- or dual-objective optimizing either both ϵ q Ir interpretation optimization trajectories. The optimizations also had an impact other properties, dissociation energies oxidative addition barriers. A similar implemented condition by solubility steric bulk.

Язык: Английский

---

Hybrid DFT Geometries and Properties for 17k Lanthanoid Complexes─The LnQM Data Set

Journal of Chemical Information and Modeling, Год журнала: 2024, Номер 64(3), С. 825 - 836

Опубликована: Янв. 18, 2024

The unique properties of lanthanoids and their diverse applications make them an indispensable part modern research industry. While the field has garnered attention, there remains a gap in available molecule data sets that facilitate both classical quantum chemistry calculations burgeoning machine learning science applications. This addresses need for comprehensive set allows comparative analysis various lanthanoids. herein presented, curated includes 17269 monolanthanoid complexes derived from 1205 distinct ligand motifs. Structures encompass all 15 +3 oxidation state exhibit molecular charges ranging −1 to +3, including structures with high spin multiplicity up 8. Starting lanthanum complexes, samples were processed permutation central lanthanoid atom, resulting highly comparable subsets, facilitating studies which influence can be investigated independently effects. provides broad range features such as PBE0-D4/def2-SVP optimized geometries optimization trajectories, while also covering ωB97M–V/def2-SVPD energies, rotational constants, dipole moments, highest occupied orbital–lowest-unoccupied orbital (HOMO–LUMO) Mulliken, Löwdin, Hirshfeld population analyses. Additionally, coordination numbers, polarizabilities, partial D4, electronegativity equilibration (EEQ), GFN2-xTB, charge extended Hückel (CEH) are included. is openly accessible may serve basis further investigations into

Язык: Английский

---

ReaLigands: A Ligand Library Cultivated from Experiment and Intended for Molecular Computational Catalyst Design

Journal of Chemical Information and Modeling, Год журнала: 2023, Номер 63(23), С. 7412 - 7422

Опубликована: Ноя. 21, 2023

Computational catalyst design requires identification of a metal and ligand that together result in the desired reaction reactivity and/or selectivity. A major impediment to translating computational designs experiments is evaluating ligands are likely be synthesized. Here, we provide solution this with our ReaLigands library contains >30,000 monodentate, bidentate (didentate), tridentate, larger cultivated by dismantling experimentally reported crystal structures. Individual from mononuclear structures were identified using modified depth-first search algorithm charge was assigned machine learning model based on quantum-chemical calculated features. In library, sorted direct ligand-to-metal atomic connections denticity. Representative principal component analysis (PCA) uniform manifold approximation projection (UMAP) analyses used analyze several tridentate categories, which revealed both diversity between categories. We also demonstrated utility implementing it building optimization tools, resulted very rapid generation barriers for 750 Rh-hydride ethylene migratory insertion.

Язык: Английский

---

Applying Large Graph Neural Networks to Predict Transition Metal Complex Energies Using the tmQM_wB97MV Data Set

Journal of Chemical Information and Modeling, Год журнала: 2023, Номер 63(24), С. 7642 - 7654

Опубликована: Дек. 4, 2023

Machine learning (ML) methods have shown promise for discovering novel catalysts but are often restricted to specific chemical domains. Generalizable ML models require large and diverse training data sets, which exist heterogeneous catalysis not homogeneous catalysis. The tmQM set, contains properties of 86,665 transition metal complexes calculated at the TPSSh/def2-SVP level density functional theory (DFT), provided a promising set catalyst systems. However, we find that trained on consistently underpredict energies chemically distinct subset data. To address this, present tmQM_wB97MV filters out several structures in found be missing hydrogens recomputes all other ωB97M-V/def2-SVPD DFT. show no pattern incorrect predictions much lower errors than those tmQM. tested were, from best worst, GemNet-T > PaiNN ≈ SpinConv SchNet. Performance improves when using only neutral instead entire set. while saturate with structures, more continue improve including charged species, indicating importance accurately capturing range oxidation states future generation model development. Furthermore, fine-tuning approach weights were initialized OC20 led drastic improvements performance, transferability between strategies

Язык: Английский

---