Deep Generative Model for the Dual-Objective Inverse Design of Metal Complexes DOI Creative Commons

Magnus Strandgaard,

Trond Linjordet, Hannes Kneiding

и другие.

Опубликована: Май 29, 2024

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 transformations. Compared to drug discovery within organic molecular space, TMCs pose further challenges including encoding chemical bonds higher complexity optimization multiple properties, a context which synthesizability is affected by additional, complex factors. In this work, we developed junction tree variational autoencoder (JT-VAE) model for generation ligands. After implementing SMILES-based metal–ligand bonds, was trained with tmQMg-L ligand library, allowing random thousands monodentate bidentate ligands full validity high novelty. The generated were labeled two target properties associated [IrL4]+ [IrL2]+ homoleptic TMCs; namely HOMO-LUMO gap (ϵ) charge (qIr), both computed at DFT level. This data used implement conditional JT-VAE generating from prompt, single or dual objective optimizing either one Y = (ϵ, qIr). Conditional able navigate central extreme regions bidimensional property interpretation based on step-wise analysis decoded trajectories.

Язык: Английский

Discovery of molybdenum based nitrogen fixation catalysts with genetic algorithms DOI Creative Commons

Magnus Strandgaard,

Julius Seumer, Jan H. Jensen

и другие.

Chemical Science, Год журнала: 2024, Номер 15(27), С. 10638 - 10650

Опубликована: Янв. 1, 2024

Using genetic algorithms and semiempirical quantum mechanical methods for discovery of nitrogen fixation catalysts.

Язык: Английский

Процитировано

3
OM-Diff: inverse-design of organometallic catalysts with guided equivariant denoising diffusion DOI Creative Commons
François Cornet, Bardi Benediktsson,

Bjarke Hastrup

и другие.

Digital Discovery, Год журнала: 2024, Номер 3(9), С. 1793 - 1811

Опубликована: Янв. 1, 2024

This work presents OM-Diff, an inverse-design framework based on a diffusion generative model for in silico design of organometallic complexes.

Язык: Английский

Процитировано

3
OM-DIFF: INVERSE-DESIGN OF ORGANOMETALLIC CATALYSTS WITH GUIDED EQUIVARIANT DENOISING DIFFUSION DOI Creative Commons
François Cornet, Bardi Benediktsson,

Bjarke Hastrup

и другие.

Опубликована: Март 28, 2024

Organometallic complexes are ubiquitous in homogeneous catalysis and other technological applications. Optimization of such for specific applications is challenging due to the large variety possible metal-ligand combinations ligand-ligand interactions. Here we present OM-Diff, an inverse design framework based on a diffusion generative model in-silico from scratch. Given importance spatial structure catalyst, directly operates all-atom (including hydrogen) representations 3D space. To handle symmetries inherent that data representation, OM-Diff combines equivariant property predictor drive sampling at inference time. The can conditionally generate novel ligands beyond those training dataset. We demonstrate potential proposed approach by designing catalysts family cross-coupling reactions, validating selection compounds with DFT calculations.

Язык: Английский

Процитировано

2
Deep Generative Model for the Dual-Objective Inverse Design of Metal Complexes DOI Creative Commons

Magnus Strandgaard,

Trond Linjordet, Hannes Kneiding

и другие.

Опубликована: Май 29, 2024

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 transformations. Compared to drug discovery within organic molecular space, TMCs pose further challenges including encoding chemical bonds higher complexity optimization multiple properties, a context which synthesizability is affected by additional, complex factors. In this work, we developed junction tree variational autoencoder (JT-VAE) model for generation ligands. After implementing SMILES-based metal–ligand bonds, was trained with tmQMg-L ligand library, allowing random thousands monodentate bidentate ligands full validity high novelty. The generated were labeled two target properties associated [IrL4]+ [IrL2]+ homoleptic TMCs; namely HOMO-LUMO gap (ϵ) charge (qIr), both computed at DFT level. This data used implement conditional JT-VAE generating from prompt, single or dual objective optimizing either one Y = (ϵ, qIr). Conditional able navigate central extreme regions bidimensional property interpretation based on step-wise analysis decoded trajectories.

Язык: Английский

Процитировано

2