IBM Achieves 210x Accuracy Boost in 12,635-Atom Protein Simulation

Researchers from Cleveland Clinic, RIKEN, and IBM have successfully simulated a 12,635-atom protein complex, marking the largest heterogeneous quantum-classical electronic-structure calculation to date and expanding the scope of quantum chemistry. The team modeled the proteins T4-Lysozyme and Trypsin, involved in immune response and digestion, binding with molecules in a water solution, achieving a 40-fold increase in system size compared to previous work. This advance focused on both scale and accuracy, demonstrating a 210-times improvement over existing approaches. “This result is one of those things you dream about,” said Dr. Kenneth Merz, PhD, lead author on the paper and leader of the Merz lab at Cleveland Clinic, suggesting a shift toward practical applications for quantum computing in complex biological systems.

Qubit Simulation Models 12,635-Atom Protein Complexes

Reaching a scale previously limited to theoretical models, researchers have successfully simulated the electronic structure of a protein complex containing 12,635 atoms using a combined quantum and classical computing approach. The simulation focused on the proteins T4-Lysozyme and Trypsin, critical components of the immune system and digestive processes, immersed in a liquid water solution, modeling their interactions with binding molecules for a more biologically relevant scenario. The team’s success wasn’t simply about increasing the size of the simulated system; it also delivered a 210-times improvement in accuracy compared to previous quantum-classical electronic-structure calculations in a specific step of the workflow. This advance was facilitated by refinements to both classical and quantum methods, utilizing two 156-qubit IBM Quantum Heron r2 processors alongside the Fugaku and Miyabi-G supercomputers.

High-performance computing experts from RIKEN played a crucial role in optimizing the workflow, which required 9,200 circuits run for over 100 hours, generating 1.3 billion measurement outcomes. Four months prior, the same group modeled the 303-atom miniprotein Trp-cage, demonstrating a 40-fold increase in system size with this latest milestone. A key innovation involved adapting a technique called wave function-based embedding, which breaks down complex calculations into smaller, manageable pieces solved by classical computers, with the most challenging segments handled by the quantum computer using sample-based quantum diagonalization. Traditional wave function-based embedding methods struggle with larger molecules due to the computational expense of defining the “fragment bath”, the highly entangled orbitals surrounding each fragment.

However, the researchers leveraged the localized nature of electron interactions within the Trypsin protein. “Information that comes from more than 7-10 angstroms away doesn’t really affect the cluster at a quantum mechanical level, in this molecule. Entanglement is already dead and gone at that distance, so one can restrict the MP2 bath expansion to a sphere centered around each atom,” explained Mario Motta, an IBM researcher and co-author of the work. This combination of algorithmic improvements and powerful hardware has positioned quantum computing as a viable tool for tackling real-world chemistry problems.

Wave Function-Based Embedding & Sample-Based Quantum Diagonalization

The pursuit of increasingly accurate molecular simulations has long been constrained by the exponential growth of computational demands as system size increases, but recent advances are challenging this limitation through a hybrid quantum-classical approach centered on wave function-based embedding and sample-based quantum diagonalization. This methodology alters how complex chemical systems are modeled, offering a pathway toward practical quantum computation for real-world problems. Researchers are now able to model molecules with unprecedented detail, moving beyond theoretical exercises to simulations with tangible relevance to biological systems and materials science. A team from Cleveland Clinic, RIKEN, and IBM recently demonstrated the power of this approach by simulating protein complexes, T4-Lysozyme and Trypsin, reaching a scale of 12,635 atoms in their largest simulation.

These proteins, critical to immune response and digestion, were modeled not in isolation, but interacting with binding molecules within a solution of water, creating a more realistic representation of their natural environment. The leap from 303 to over 12,000 atoms represents a 40-fold increase in system size, and the improvements extend beyond sheer scale. The core innovation lies in how the computational burden is distributed. Wave function-based embedding fragments the molecule into smaller, manageable “clusters” solved by classical computers. The most complex clusters, where entanglement between atoms is strongest, are then tackled by a quantum computer using sample-based quantum diagonalization, which samples the vast space of possible electronic configurations, identifying key configurations for the classical computer to refine. However, the team refined wave function-based embedding to focus on localized electron interactions, recognizing that information from beyond a certain distance (approximately 7-10 angstroms) has minimal impact on the cluster’s quantum mechanical behavior.

Dr. Kenneth Merz said he’s seen dramatic improvements in the ability of computers to model chemistry in his career. This simplification allowed them to scale the simulation to an unprecedented size. The team utilized up to 94 qubits across two IBM Quantum Heron r2 processors, running 9,200 circuits for over 100 hours and collecting 1.3 billion measurement outcomes, making this the most resource-intensive quantum chemistry execution to date. Merz believes this workflow will soon outperform classical methods, potentially accelerating advancements in pharmaceutical development, materials science, and broader chemistry research, ultimately leading to “better lifesaving drugs, faster” and “better materials for the technology in your home or for national infrastructure.”

40x System Size & 210x Accuracy Improvements Achieved

Cleveland Clinic researchers, in collaboration with RIKEN and IBM, are pushing the boundaries of molecular simulation with a newly demonstrated quantum-centric supercomputing workflow. The team successfully modeled a protein complex containing 12,635 atoms, the largest heterogeneous quantum-classical electronic-structure calculation to date, a feat previously considered unattainable. This achievement builds upon earlier work, notably the modeling of the 303-atom miniprotein Trp-cage just four months prior, and signifies a substantial leap in both scale and reliability for quantum chemistry simulations. Unlike the Trp-cage simulation, which focused on the isolated protein, this latest work modeled the proteins interacting with binding molecules within a liquid water solution, creating a more realistic and computationally demanding scenario. Achieving this required more than simply increasing computational power; the researchers refined both classical and quantum methods employed in the workflow.

The initial Trp-cage result on which this new work builds relies on a technique called wave function-based embedding, which fragments the calculation into computationally tractable pieces called “clusters.” Classical computers solve the simpler clusters, and then a quantum computer uses sample-based quantum diagonalization to solve the more complex clusters, those involving more entanglement between atoms in the miniprotein. The classical computers then stitch the molecule back together. Merz also noted the slowing pace of improvement in classical computing, suggesting that quantum computing may be essential for achieving further substantial advances in the field.

High-performance computing experts from RIKEN were instrumental in optimizing the process, ensuring efficient data processing and workflow integration. A key innovation enabling this scale was a wave function-based embedding technique, which fragments complex calculations into manageable pieces. Classical computers handle simpler fragments, while a quantum computer, employing sample-based quantum diagonalization, tackles the more intricate portions involving significant entanglement between atoms, and the classical computers then reassemble the complete molecule.