NMR is a powerful tool for identifying molecules, widely used in chemistry and the pharmaceutical industry for drug discovery and quality control. However, the manual interpretation of NMR spectra, particularly in the case of mixtures, poses significant challenges. This complexity is even more difficult to manage when using low-field NMR, which is the more cost-effective option. The advantage of analysing mixtures with compact, low-field NMR instruments will enable the application of NMR in personalised medicine and beyond.
Our project aims to revolutionize Nuclear Magnetic Resonance (NMR) analysis, enhancing its application in diagnostics and personalized medicine. By leveraging advanced quantum simulations, we seek to improve the speed and accuracy of NMR analysis, for both high-field and low-field devices. This will make NMR analysis more accessible by broadening the usability of more affordable low-field NMR devices. In practice, the majority of samples encountered in industrial and clinical settings are mixtures, not clean single-component compounds. Mixture analysis is central to high-value workflows such as: quality control, forensic analytics, environmental control, industrial analytics and metabolomics, where understanding complex biological mixtures, such as blood and urine, ultimately paves the way for individualized therapies.
In previous grant-funded projects, we have developed a spectrum prediction workflow that calculates NMR parameters and generates spectra for various spectrometer frequencies. In the EIC Transition project, we aim to refine the parameter calculation step to enhance precision across diverse molecular structures. A key component of our approach is the development of an advanced mixture analysis tool that will assist in analyzing biological samples. As we progress, we will incorporate machine learning-driven structure identification, complemented by a verification step that compares with ab-initio predictions. This dual approach ensures high accuracy and explainability, addressing the limitations of existing AI solutions in the NMR prediction space.
The software’s ability to enhance low-field NMR devices, making them suitable for complex analyses, makes NMR a viable option for advanced analytics, particularly in cost-sensitive sectors such as pharmaceuticals, biotechnology, and environmental science.
