Where Quantum Computing Meets GPU-accelerated Supercomputing
Accurately simulating biomolecular systems at scale is essential for understanding the molecular mechanisms that underpin biology, health and disease. To address this challenge, a research effort between University College London, Technical University of Munich, Ludwig-Maximilians-Universität München (LMU), NVIDIA, Leibniz Supercomputing Centre (LRZ), QMatter and IQM Quantum Computers has developed an integrated biomolecular simulation pipeline that couples GPU acceleration, conventional supercomputing and quantum computing using the NVIDIA CUDA-Q platform. By integrating quantum computing into large-scale accelerated simulation workflows, the pipeline allows researchers to target the most challenging molecular effects while retaining the performance and scalability of modern supercomputing.
The research project follows a multiscale simulation approach that applies different levels of accuracy across a biomolecular system. Quantum computing models the chemically active regions of interest, while established computational chemistry and classical simulation methods describe the surrounding environment. This combination preserves quantum-level detail for biochemical processes within their full biological context, from distant regions of the molecule to membranes and solvent. As an initial demonstration, the pipeline was applied to a G-protein-coupled receptor, a key class of membrane proteins in cellular signaling.
G-protein-coupled receptors (GPCRs) are essential molecular gatekeepers that allow signals, hormones, and neurotransmitters to pass information across cell membranes. With more than 800 identified subtypes, GPCRs control critical physiological processes ranging from heart function to brain signaling and are arguably the most important class of drug targets, with roughly one-third of approved medicines acting on them. Their size, structural complexity, and membrane environment make them particularly challenging to model, especially given the long timescales of their signaling mechanisms. The multiscale, quantum-accelerated approach developed in this project opens new possibilities for studying these biologically and pharmacologically important systems.
Realizing this workflow at scale requires tight integration between quantum computing and GPU-accelerated supercomputing. The quantum component was executed on Euro-Q-Exa, an IQM 54-qubit system hosted at LRZ, which involves evolving a quantum state through time to search for important configurations of electrons arranged in molecular orbitals. These configurations are subsequently interacted to evaluate their energy, a workload that is highly amenable to GPU acceleration. In this project, we utilized NVIDIA Hopper GPUs on the AI cluster hosted at LRZ, which is part of BayernKI, Bavaria’s academic AI research infrastructure. The postprocessing element of the work was further scaled to a billion quantum configurations using 1,200 NVIDIA H100 GPUs on the Eos cluster.
Credit: LRZ / V. Hohenegger
The GPU-accelerated post-processing demonstrates strong scaling, paving the way for an efficient scale-up of the workflow on future AI Factories. Future studies will also build on the developed techniques for coordinating heterogeneous computing resources in large-scale biomolecular simulation, enabling iterative hybrid workflows across complementary architectures. Together, this approach expands the role of computation in biomolecular science, enabling deeper insights into molecular mechanisms under realistic conditions.
See the announcements from our collaborators:
For further information please contact hello@qmatter.xyz.