Cleveland Clinic scientists harness hybrid quantum-classical computing to accurately predict proton affinity—paving the way for revolutionary advances in computational chemistry and molecular modeling.
Research: Integrating Machine Learning and Quantum Circuits for Proton Affinity Predictions. Image Credit: Corona Borealis Studio / Shutterstock
Kenneth Merz, PhD, of the Cleveland Clinic's Center for Computational Life Sciences, and a research team are testing quantum computing's abilities in chemistry by integrating machine learning and quantum circuits.
Chemistry is one of the areas where quantum computing shows the most potential because the technology can predict an unlimited number of possible outcomes. To determine quantum computing's ability to perform complex chemical calculations, Dr. Merz and Hongni Jin, PhD, decided to test its ability to simulate proton affinity, a fundamental chemical process critical to life.
Dr. Merz and Dr. Jin focused on using machine learning applications on quantum hardware. This is a critical advantage over other quantum research, which relies on simulators to mimic a quantum computer's abilities. In this study, published in the Journal of Chemical Theory and Computation, the team demonstrated the capabilities of quantum machine learning by creating a model that was able to predict proton affinity more accurately than classical computing.
Quantum computing is an entirely new method of computing that operates differently than classical computers. Classical computers depend on bits, a series of 1s and 0s, to solve problems. A quantum computer uses qubits, which can exist in multiple states simultaneously and are not limited to 1s or 0s.
Quantum Computing 101
When classical computers solve complex problems, bits are put through logic gates. Quantum gates facilitate qubits, which act in a way that is impossible on classical computers. Quantum gates allow qubits to exist in multiple states, allowing them to test all the "rules" put in place by gates and all the potential outcomes simultaneously. This is essential in chemistry, where molecules can behave in ways that have unlimited possible outcomes.
To narrow the scope of the study, the team chose to focus on proton affinity in the gas phase. Proton affinity is the ability of a molecule to attract and hold a proton. This process is a critical chemical endpoint that is challenging to study in the gas phase because most compounds do not easily evaporate and can be destroyed by heat, limiting the ability to carry out experiments. Dr. Merz says these experiments are time-consuming and can only be applied to small or medium-sized molecules, making the problem an ideal test for quantum computing.
For this project, the team applied machine learning and quantum circuits created using quantum gates. Dr. Jin says the QML model they designed was trained on 186 different factors. The research team compared the model's accuracy for predicting proton affinity between the classical computer and the hybrid quantum and classical computing method.
"This project was one of our first experiences with QML," Dr. Merz says. "Machine learning has already proven to be useful in chemistry because of its ability to correlate chemical structures with their physical-chemical properties and predict reaction outcomes. With the power of quantum computing, it can surpass even the most advanced supercomputer with its compute power."
Source:
Journal reference:
- Integrating Machine Learning and Quantum Circuits for Proton Affinity Predictions Hongni Jin and Kenneth M. Merz Jr Journal of Chemical Theory and Computation 2025 21 (5), 2235-2243 DOI: 10.1021/acs.jctc.4c01609, https://pubs.acs.org/doi/10.1021/acs.jctc.4c01609