hanneskn

• Sandonas, Leonardo Medrano; Badia, David Balcells; Barnowsky, Tom; Bochkarev, Anton; Cole, Jacqueline & Deringer, Volker L. [Vis alle 24 forfattere av denne artikkelen] (2026). Towards sustainable exploration of chemical spaces with machine learning. Digital Discovery. doi: 10.1039/d6dd00157b. Fulltekst i vitenarkiv Vis sammendrag

  Artificial intelligence is transforming molecular and materials science, but its growing computational and data demands raise critical sustainability challenges. In this perspective, we examine resource considerations across the AI-driven discovery pipeline–from quantum-mechanical (QM) data generation and model training to automated, self-driving research workflows–building on discussions from the “SusML workshop: towards sustainable exploration of chemical spaces with machine learning” held in Dresden, Germany. In this context, the availability of large quantum datasets has enabled rigorous benchmarking and rapid methodological progress, while also incurring substantial energy and infrastructure costs. We highlight emerging strategies to enhance efficiency, including general-purpose machine learning (ML) models, multi-fidelity approaches, model distillation, and active learning. Moreover, incorporating physics-based constraints within hierarchical workflows, where fast ML surrogates are applied broadly and high-accuracy QM methods are used selectively, can further optimize resource use without compromising reliability. Equally important is bridging the gap between idealized computational predictions and real-world conditions by accounting for synthesizability and multi-objective design criteria, which is essential for practical impact. Finally, we argue that sustainable progress will rely on open data and models, reusable workflows, and domain-specific AI systems that maximize scientific value per unit of computation, enabling efficient and responsible discovery of technological materials and therapeutics.

• Kneiding, Hannes & Badia, David Balcells (2026). Δ-Machine Learning for the Prediction of Metal Complex Properties. Chemistry - A European Journal. ISSN 0947-6539. doi: 10.1002/chem.71487. Fulltekst i vitenarkiv
• Kneiding, Hannes & Badia, David Balcells (2025). tmQMg* Data Set: Excited State Properties of 74k Transition Metal Complexes. Journal of Chemical Information and Modeling. ISSN 1549-9596. 65(21), s. 11766–11777. doi: 10.1021/acs.jcim.5c01958. Fulltekst i vitenarkiv Vis sammendrag

  The application of machine learning approaches to meaningful problems in chemistry and materials science is still challenged by the limited availability of data. In order to close this gap, we report the tmQMg* data set, which provides excited state properties for 74k mononuclear transition metal complexes extracted from the Cambridge Structural Database. All properties were computed at the TD-DFT ωB97xd/def2SVP level of theory. The strongest electron excitations in the ultraviolet, visible, and near-infrared ranges are included, together with the wavelengths and intensities of the first 30 excited states. Further, natural transition orbitals were computed for the strongest excitations in the visible range to determine the nature of the associated charge transfers. By computing the TD-DFT spectra in both gas phase and acetone, we quantified solvatochromic effects, which are also provided with the data set, in terms of both wavelength shifts and intensity changes. The tmQMg* data set will enable the development of discriminative and generative artificial intelligence models with respect to absorption spectra, charge transfer character, and solvatochromism, enabling novel advances in the field of transition metal photochemistry.

• Strandgaard, Magnus; Linjordet, Trond; Kneiding, Hannes; Burnage, Arron Luc; Nova, Ainara & Jensen, Jan Halborg [Vis alle 7 forfattere av denne artikkelen] (2025). A Deep Generative Model for the Inverse Design of Transition Metal Ligands and Complexes. JACS Au. 5(5), s. 2294–2308. doi: 10.1021/jacsau.5c00242. Fulltekst i vitenarkiv Vis sammendrag

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

• Burnage, Arron Luc; Kneiding, Hannes; Nova, Ainara & Badia, David Balcells (2025). Cross-coupling catalysts for inactive substrates from a three-dimensional virtual screening. Journal of Catalysis. ISSN 0021-9517. 452. doi: 10.1016/j.jcat.2025.116446. Fulltekst i vitenarkiv Vis sammendrag

  We report a three-dimensional virtual screening method to search a chemical space formed of ligands from the tmQMg-L library for Pd Suzuki–Miyaura cross-coupling catalysts that has future potential to be applied to a broader range of catalytic reactions. It utilises three criteria related to important steps in the catalytic cycle of the reaction to distinguish between candidate Pd(0) species, namely the oxidative addition and reductive elimination barriers and the binding energy of acetonitrile. This method ultimately leads to identifying a region of chemical space that provides a good compromise between the three search criteria, highlighting a smaller set of ligands that merits further exploration. After the screening, several candidates are confirmed as catalysts based on literature searches. We also found Pd(0) complexes with high potential in catalytic cross-couplings involving less reactive substrates that require the oxidative addition of strong C–O bonds.

Se alle arbeider i NVA

• González, Lucía Morán; Kneiding, Hannes; Badia, David Balcells & Flores, Ainara Nova (2026). High-Throughput Computational Screening of Bifunctional Catalysts for CO₂ Hydrogenation. doi: https:/ccsc2026.github.io/. Fulltekst i vitenarkiv
• Kneiding, Hannes & Badia, David Balcells (2025). tmQMg* dataset: excited state properties of 74k transition metal complexes. Fulltekst i vitenarkiv
• Kneiding, Hannes & Badia, David Balcells (2025). Graphs for machine learning applications in chemistry. Fulltekst i vitenarkiv
• Kneiding, Hannes; Badia, David Balcells & Flores, Ainara Nova (2024). Directional Multiobjective Optimization of Metal Complexes In Vast Chemical Spaces. Fulltekst i vitenarkiv Vis sammendrag

  Transition metal complexes (TMCs) are crucial for a multitude of applications including nanoporous materials, renewable energies, and medicinal chemistry. Their development time can be drastically reduced by identifying promising candidates in silico before validating hits with experiments. However, sufficiently large and diverse ligand libraries that enable the efficient computational optimization of TMCs in vast chemical spaces are still lacking. Here, we present our recent work on the open tmQMg-L [1] dataset including 30K experimentally relevant TMC ligands extracted from the Cambridge Structural Database. All ligands come with robustly defined charge and metal-coordination mode. We demonstrate how the tmQMg-L dataset can be leveraged in the multiobjective optimization of TMCs with respect to different properties of interest. In particular, we employ a genetic algorithm with full-ligand genetic operations and multiobjective selection rules to simultaneously maximize the HOMO-LUMO gap and polarizability of TMCs in chemical spaces based on a palladium(II) scaffold with up to 1B unique members. The final hits occupy, or are very close, to the Pareto front and their aim and scope can be controlled using our newly introduced Pareto-Lighthouse functionality. Furthermore, we discuss how the ligands in tmQMg-L can be used to train a variational autoencoder in order to generate novel ligands ready for their use in evolutionary optimization approaches. [1] Kneiding, H., Nova, A., Balcells, D. Nat. Comput. Sci. 2024, accepted

• Kneiding, Hannes & Badia, David Balcells (2024). Quantum-informed graph representations for the prediction of molecular properties of transition metal complexes. Fulltekst i vitenarkiv Vis sammendrag

  Transition metal complexes (TMCs) are essential for a wide range of applications including renewable energies, medicinal chemistry, and nanoporous materials. Their investigation with high-throughput screenings for properties of interest using computational methods is still associated with prohibitively large computational costs rendering them impractical. Predictive machine learning (ML) can close this gap and enable the exploration of vast chemical spaces including millions of compounds. However, due to the complex chemistry present in TMCs, the construction of sufficiently accurate ML models is still impeded and an area of current research. Here, we present work in which we devised highly expressive graph representations of TMCs based on quantum chemistry information and used them in conjunction with graph neural networks in order to facilitate the prediction of quantum properties of interest. [1] In particular, these graph representations are based on Natural Bond Orbital [2] analysis data that provides a rich feature set to be leveraged in ML approaches leading. We show how they outperform methods from the literature based solely on structural features and furthermore demonstrate their application in Δ-learning approaches, in which properties are first calculated at a low level of theory to then predict the corresponding corrections to the target values at a high level of theory. Quantum information extracted from the necessary low level calculations will be used to featurize the TMC graph representations. [1] KNEIDING, Hannes, et al. Deep learning metal complex properties with natural quantum graphs. Digital Discovery, 2023, 2. Jg., Nr. 3, S. 618-633. [2] GLENDENING, Eric D.; LANDIS, Clark R.; WEINHOLD, Frank. NBO 7.0: New vistas in localized and delocalized chemical bonding theory. Journal of computational chemistry, 2019, 40.25: 2234-2241.

Se alle arbeider i NVA