Photo: © Ulrich Zinell
Heid Lab · TU Wien · Institute of Materials Chemistry
Deep learning for chemical reactions
Bridging AI and mechanistic chemistry to enable high-throughput screening of chemical reactions.
What we do
Accelerating chemical discovery with AI
The Heid Lab is led by Esther Heid, Assistant Professor for Machine Learning for Sustainable Chemistry at TU Wien. We develop machine learning methods to predict and model chemical reactions, ranging from barrier heights and reaction energies to transition state geometries. Our work combines graph neural networks, generative models, and quantum chemistry.
Our research spans reaction property prediction, transition state modeling, and ML-accelerated automated reaction pathway search across gas-phase, organocatalytic, and biocatalytic reactions.
Explore our work →Projects
Funded by the ERC and FWF
Our work is supported by multiple grants including a highly competitive ERC Starting Grant and an FWF START Prize, supporting the group's long-term research on machine learning for chemical reactions as part of the DeepRxn project.
FWF START Prize
What we've been up to
News
-
Our Seminar Series "Interdisciplinary Foundations of Deep Learning in Chemistry" opens its doors
We are proud to announce that Esther and Max will host the Seminar Series Interdisciplinary Foundations of Deep Learning in Chemistry, which is now also accessible to an interested audience as well as to motivated students in mathematics, data science, informatics, computer science, computational science and engineering, physics, chemistry, and chemical engineering. -
Florentin joins for his master's thesis
We are happy to welcome Florentin again, who will write his master's thesis with Jasper. -
Max was interviewed by Kurier about his Ph.D. project
In the Austrian daily Kurier, Max talks about his DOC-funded Ph.D. project. "The goal is not to collect every recipe in the world, but to understand the principles behind them". Read the full article here (in German).
Selected work
Featured research
Here's a selection of recent highlights.
GoFlow: transition-state geometry prediction with flow matching and E(3)-equivariant networks
L. Galustian, K. Mark, J. Karwounopoulos, M. P.-P. Kovar, E. Heid
Read paper →Graph-based prediction of reaction barrier heights with on-the-fly transition states
J. Karwounopoulos, J. De Landsheere, L. Galustian, T. Jechtl, E. Heid
Read paper →ChemTorch: a deep learning framework for reaction property prediction
J. De Landsheere, A. Zamyatin, J. Karwounopoulos, E. Heid
Read paper →Contact
Want to get involved?
Whether you're thinking about joining, looking to collaborate, or just want to exchange ideas — we'd be glad to hear from you.