Johannes Karwounopoulos
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Johannes Karwounopoulos has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Esther Heid
Institute of Materials Chemistry, TU Wien, A-1060 Vienna, Austria
11 shared publications Leonard Galustian
Institute of Materials Chemistry, TU Wien, A-1060 Vienna, Austria
4 shared publications- Stefan Boresch
University of Vienna
4 shared publications - Anna Katharina Picha
University of Vienna
3 shared publications - Jasper De Landsheere
Institute of Materials Chemistry, TU Wien, A-1060 Vienna, Austria
3 shared publications Konstantin Mark
Institute of Materials Chemistry, TU Wien, A-1060 Vienna, Austria
3 shared publications
Latest Publications
- Open Access
- 14 October 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (10) , 4499-4511
moTSart: accelerating automated transition state search with generative models in a low-data regime
We introduce moTSart, a framework for generative AI accelerated, automated transition state search.
- 14 October 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (10) , 4499-4511
Open AccessmoTSart: accelerating automated transition state search with generative models in a low-data regime
We introduce moTSart, a framework for generative AI accelerated, automated transition state search.
- Open Access
- 3 September 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (18) , 9274-9289
GRACE-OFF: A Machine-Learned Interatomic Potential for Organic Liquids Using the GRACE Architecture
Machine-learned interatomic potentials (MLIPs) have become an increasingly important tool for molecular dynamics (MD) simulations, enabling near quantum-mechanical accuracy at significantly reduced computational cost. Recent studies indicate that the Graph Atomic Cluster Expansion (GRACE) neural network architecture delivers strong performance in materials chemistry. In this work, we assess the GRACE architecture for the prediction of potential energy surfaces for organic molecules and introduce GRACE-OFF (GRACE Organic Force Field). GRACE models of varying depth (one-layer and two-layer) and size (small, medium, large) are trained on the SPICE v2.0 data set. We validate the resulting models using a variety of benchmarks. These include single-point energy and force predictions, torsional energy profiles, condensed phase properties of organic liquids and water (thermodynamic properties, self-diffusion coefficients, radial distribution functions, and temperature-dependent water density), as well as the stability of biomolecular MD simulations for gas-phase Ala15 and solvated crambin. For the single-molecule benchmarks (single point energies and forces, torsional energy profiles), the one-layer models showed only mediocre performance, whereas the two-layer models outperformed the MACE-OFF models to which we compare. For the condensed phase properties, the two-layer models gave consistently better results than the MACE-OFF family of MLIPs. For water and hexane, the GRACE-OFF models also beat the much more expensive small UMA/OMol25 (S) model. The two-layer GRACE-OFF models accurately reproduce experimental water radial distribution functions and predict water densities in close agreement with experimental data over a temperature range from 270 to 330 K. Benchmarks demonstrate that GRACE-OFF achieves higher MD performance than comparable MACE-OFF models in both single and double precision. This establishes GRACE-OFF as an accurate and computationally efficient foundation potential for routine simulations of organic liquids and biomolecular systems.
- 3 September 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (18) , 9274-9289
Open AccessGRACE-OFF: A Machine-Learned Interatomic Potential for Organic Liquids Using the GRACE Architecture
Machine-learned interatomic potentials (MLIPs) have become an increasingly important tool for molecular dynamics (MD) simulations, enabling near quantum-mechanical accuracy at significantly reduced computational cost. Recent studies indicate that the Graph Atomic Cluster Expansion (GRACE) neural network architecture delivers strong performance in materials chemistry. In this work, we assess the GRACE architecture for the prediction of potential energy surfaces for organic molecules and introduce GRACE-OFF (GRACE Organic Force Field). GRACE models of varying depth (one-layer and two-layer) and size (small, medium, large) are trained on the SPICE v2.0 data set. We validate the resulting models using a variety of benchmarks. These include single-point energy and force predictions, torsional energy profiles, condensed phase properties of organic liquids and water (thermodynamic properties, self-diffusion coefficients, radial distribution functions, and temperature-dependent water density), as well as the stability of biomolecular MD simulations for gas-phase Ala15 and solvated crambin. For the single-molecule benchmarks (single point energies and forces, torsional energy profiles), the one-layer models showed only mediocre performance, whereas the two-layer models outperformed the MACE-OFF models to which we compare. For the condensed phase properties, the two-layer models gave consistently better results than the MACE-OFF family of MLIPs. For water and hexane, the GRACE-OFF models also beat the much more expensive small UMA/OMol25 (S) model. The two-layer GRACE-OFF models accurately reproduce experimental water radial distribution functions and predict water densities in close agreement with experimental data over a temperature range from 270 to 330 K. Benchmarks demonstrate that GRACE-OFF achieves higher MD performance than comparable MACE-OFF models in both single and double precision. This establishes GRACE-OFF as an accurate and computationally efficient foundation potential for routine simulations of organic liquids and biomolecular systems.
- Open Access
- 18 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
GRACE-OFF: A machine-learned interatomic potential for organic liquids using the GRACE architecture
- 18 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessGRACE-OFF: A machine-learned interatomic potential for organic liquids using the GRACE architecture
- Open Access
- 1 June 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Toward On-the-Fly Prediction of Reaction Energetics for High-Throughput Screening
Machine learning is increasingly used to predict reaction properties such as barrier heights, reaction energies, rates, or yields, as well as the underlying molecular geometries, including transition state structures. While such predictions have the potential to provide mechanistic insight for high-impact applications such as synthesis planning and reaction optimization, the field remains at an early stage of development. This Perspective discusses and critically assesses the current state-of-the art in (organic) reaction property prediction, highlighting the key limitations related to data availability and quality, molecular and transformation representations, and machine learning architectures used in both predictive and generative models. A special focus is given on current challenges and on possible paths forward toward efficient and accurate machine learning models for on-the-fly prediction of reaction energetics.
- 1 June 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessToward On-the-Fly Prediction of Reaction Energetics for High-Throughput Screening
Machine learning is increasingly used to predict reaction properties such as barrier heights, reaction energies, rates, or yields, as well as the underlying molecular geometries, including transition state structures. While such predictions have the potential to provide mechanistic insight for high-impact applications such as synthesis planning and reaction optimization, the field remains at an early stage of development. This Perspective discusses and critically assesses the current state-of-the art in (organic) reaction property prediction, highlighting the key limitations related to data availability and quality, molecular and transformation representations, and machine learning architectures used in both predictive and generative models. A special focus is given on current challenges and on possible paths forward toward efficient and accurate machine learning models for on-the-fly prediction of reaction energetics.
- Open Access
- 2 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
GRACE-OFF: A machine-learned interatomic potential for organic liquids using the GRACE architecture
Machine-learned interatomic potentials (MLIPs) have become an increasingly important tool for molecular dynamics (MD) simulations, enabling near quantum-mechanical accuracy at significantly reduced computational cost. In this work, we assess the Graph Atomic Cluster Expansion (GRACE) neural network architecture for the prediction of potential energy surfaces. GRACE models of varying depth and size are trained on the SPICE v2.0 dataset—which comprises a large and chemically diverse set of organic molecules—and are integrated into the MD program ASE. We validate the resulting models using a variety of benchmarks. These include single-point energy and force predictions, torsional energy profiles, condensed phase properties of organic liquids and water (thermodynamic properties, self-diffusion coefficients, radial distribution functions, and temperature-dependent water density), as well as the stability of biomolecular MD simulations for gas-phase Ala 15 and solvated crambin. For the single molecule benchmarks (single point energies and forces, torsional energy profiles) the one-layer models showed only mediocre performance, whereas the two-layer models outperformed the MACE-OFF models to which we compare to, including the latest MACE-OFF24(M). For the condensed phase properties, in particular the medium and large two-layer models gave significantly better results than the MACE-OFF family of MLIPs. For water and hexane, the GRACE-OFF models also beat the much more expensive UMA(S) model. In particular, the two-layer GRACE-OFF models accurately reproduce experimental water radial distribution functions and predict water densities in close agreement with experimental data over a wide temperature range, substantially reducing the systematic overestimation observed for MACE-OFF. Finally, GPU throughput benchmarks demonstrate that GRACE-OFF achieves substantially higher MD performance than comparable MACE models in both single and double precision. This establishes GRACE-OFF as an accurate and computationally efficient foundation potential for routine simulations of organic liquids and biomolecular systems.
- 2 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessGRACE-OFF: A machine-learned interatomic potential for organic liquids using the GRACE architecture
Machine-learned interatomic potentials (MLIPs) have become an increasingly important tool for molecular dynamics (MD) simulations, enabling near quantum-mechanical accuracy at significantly reduced computational cost. In this work, we assess the Graph Atomic Cluster Expansion (GRACE) neural network architecture for the prediction of potential energy surfaces. GRACE models of varying depth and size are trained on the SPICE v2.0 dataset—which comprises a large and chemically diverse set of organic molecules—and are integrated into the MD program ASE. We validate the resulting models using a variety of benchmarks. These include single-point energy and force predictions, torsional energy profiles, condensed phase properties of organic liquids and water (thermodynamic properties, self-diffusion coefficients, radial distribution functions, and temperature-dependent water density), as well as the stability of biomolecular MD simulations for gas-phase Ala 15 and solvated crambin. For the single molecule benchmarks (single point energies and forces, torsional energy profiles) the one-layer models showed only mediocre performance, whereas the two-layer models outperformed the MACE-OFF models to which we compare to, including the latest MACE-OFF24(M). For the condensed phase properties, in particular the medium and large two-layer models gave significantly better results than the MACE-OFF family of MLIPs. For water and hexane, the GRACE-OFF models also beat the much more expensive UMA(S) model. In particular, the two-layer GRACE-OFF models accurately reproduce experimental water radial distribution functions and predict water densities in close agreement with experimental data over a wide temperature range, substantially reducing the systematic overestimation observed for MACE-OFF. Finally, GPU throughput benchmarks demonstrate that GRACE-OFF achieves substantially higher MD performance than comparable MACE models in both single and double precision. This establishes GRACE-OFF as an accurate and computationally efficient foundation potential for routine simulations of organic liquids and biomolecular systems.
- Open Access
- 9 January 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Toward On-the-Fly Prediction of Reaction Energetics for High-Throughput Screening
Machine learning is increasingly used to predict reaction properties such as barrier heights, reaction energies, rates, or yields, as well as the underlying molecular geometries, including transition state structures. While such predictions have the potential to provide mechanistic insight for high-impact applications such as synthesis planning and reaction optimization, the field remains at an early stage of development. This Perspective discusses and critically assesses the current state-of-the art in reaction property prediction, highlighting the key limitations related to data availability and quality, molecular and transformation representations, and machine learning architectures used in both predictive and generative models. A special focus is given on current challenges and on possible paths forward toward efficient and accurate machine learning models for on-the-fly prediction of reaction energetics.
- 9 January 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessToward On-the-Fly Prediction of Reaction Energetics for High-Throughput Screening
Machine learning is increasingly used to predict reaction properties such as barrier heights, reaction energies, rates, or yields, as well as the underlying molecular geometries, including transition state structures. While such predictions have the potential to provide mechanistic insight for high-impact applications such as synthesis planning and reaction optimization, the field remains at an early stage of development. This Perspective discusses and critically assesses the current state-of-the art in reaction property prediction, highlighting the key limitations related to data availability and quality, molecular and transformation representations, and machine learning architectures used in both predictive and generative models. A special focus is given on current challenges and on possible paths forward toward efficient and accurate machine learning models for on-the-fly prediction of reaction energetics.
- Open Access
- 4 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
ChemTorch: A Deep Learning Framework for Benchmarking and Developing Chemical Reaction Property Prediction Models
Modeling of chemical reactions is essential for understanding kinetic mechanisms and predicting possible outcomes of reacting systems. Quantum mechanical calculations are accurate but often prohibitively expensive. Deep learning has emerged as a faster alternative, but progress is slowed by a fragmented software ecosystem that hinders reuse, fair comparison, and reproducibility. We present ChemTorch, an open-source framework that streamlines model development, experimentation, hyperparameter tuning, and benchmarking through modular pipelines, standardized configuration, and built-in data splitters for in- and out-of-distribution evaluation. We envision ChemTorch as a foundation for community-driven method development and reproducible benchmarking in chemical reaction modeling. As a first step toward unified benchmarks, we compare four representative modalities for barrier-height prediction on the RDB7 dataset, including fingerprint-, sequence-, graph-, and 3D-based approaches. Our results highlight clear advantages of structurally informed models and sharp performance drops under out-of-distribution conditions, highlighting the importance of rigorous benchmarking.
- 4 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessChemTorch: A Deep Learning Framework for Benchmarking and Developing Chemical Reaction Property Prediction Models
Modeling of chemical reactions is essential for understanding kinetic mechanisms and predicting possible outcomes of reacting systems. Quantum mechanical calculations are accurate but often prohibitively expensive. Deep learning has emerged as a faster alternative, but progress is slowed by a fragmented software ecosystem that hinders reuse, fair comparison, and reproducibility. We present ChemTorch, an open-source framework that streamlines model development, experimentation, hyperparameter tuning, and benchmarking through modular pipelines, standardized configuration, and built-in data splitters for in- and out-of-distribution evaluation. We envision ChemTorch as a foundation for community-driven method development and reproducible benchmarking in chemical reaction modeling. As a first step toward unified benchmarks, we compare four representative modalities for barrier-height prediction on the RDB7 dataset, including fingerprint-, sequence-, graph-, and 3D-based approaches. Our results highlight clear advantages of structurally informed models and sharp performance drops under out-of-distribution conditions, highlighting the importance of rigorous benchmarking.
- Open Access
- 21 October 2025
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 4 (12) , 3492-3501
GoFlow: efficient transition state geometry prediction with flow matching and E(3)-equivariant neural networks
We introduce GoFlow, a package to generate 3D transition state geometries using flow matching.
- 21 October 2025
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 4 (12) , 3492-3501
Open AccessGoFlow: efficient transition state geometry prediction with flow matching and E(3)-equivariant neural networks
We introduce GoFlow, a package to generate 3D transition state geometries using flow matching.
- Open Access
- 8 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
GoFlow: Efficient Transition State Geometry Prediction with Flow Matching and E(3)-Equivariant Neural Networks
Transition state (TS) geometries of chemical reactions are key to understanding reaction mechanisms and estimating kinetic properties. Inferring these directly from 2D reaction graphs offers chemists a powerful tool for rapid and accessible reaction analysis. Quantum chemical methods for computing TSs are computationally intensive and often infeasible for larger molecular systems. Recently, deep learning–based diffusion models have shown promise in generating TSs from 2D reaction graphs for single-step reactions. However, framing TS generation as a diffusion process, by design, requires a prohibitively large number of sampling steps during inference. Here we show that modeling TS generation as an optimal transport flow problem, solved via E(3)-equivariant flow matching with geometric tensor networks, achieves over a hundredfold speedup in inference while improving geometric accuracy compared to the state-of-the-art. This breakthrough increase in sampling efficiency and predictive accuracy enables the practical use of deep learning-based TS generators in high-throughput settings for larger and more complex chemical systems. Our method, GoFlow, thus represents a significant methodological advancement in machine learning-based TS generation, bringing it closer to widespread use in computational chemistry workflows.
- 8 October 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessGoFlow: Efficient Transition State Geometry Prediction with Flow Matching and E(3)-Equivariant Neural Networks
Transition state (TS) geometries of chemical reactions are key to understanding reaction mechanisms and estimating kinetic properties. Inferring these directly from 2D reaction graphs offers chemists a powerful tool for rapid and accessible reaction analysis. Quantum chemical methods for computing TSs are computationally intensive and often infeasible for larger molecular systems. Recently, deep learning–based diffusion models have shown promise in generating TSs from 2D reaction graphs for single-step reactions. However, framing TS generation as a diffusion process, by design, requires a prohibitively large number of sampling steps during inference. Here we show that modeling TS generation as an optimal transport flow problem, solved via E(3)-equivariant flow matching with geometric tensor networks, achieves over a hundredfold speedup in inference while improving geometric accuracy compared to the state-of-the-art. This breakthrough increase in sampling efficiency and predictive accuracy enables the practical use of deep learning-based TS generators in high-throughput settings for larger and more complex chemical systems. Our method, GoFlow, thus represents a significant methodological advancement in machine learning-based TS generation, bringing it closer to widespread use in computational chemistry workflows.
- Open Access
- 15 September 2025
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 4 (11) , 3208-3216
Graph-based prediction of reaction barrier heights with on-the-fly prediction of transition states
The accurate prediction of reaction barrier heights is crucial for understanding chemical reactivity and guiding reaction design.
- 15 September 2025
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 4 (11) , 3208-3216
Open AccessGraph-based prediction of reaction barrier heights with on-the-fly prediction of transition states
The accurate prediction of reaction barrier heights is crucial for understanding chemical reactivity and guiding reaction design.