Next-Gen NMR Prediction and Analysis: A Cutting-Edge Tool for High-Precision Structure Identification

Informations projet

HQS-NextNMR

N° de convention de subvention: 101214526

DOI

10.3030/101214526

Date de signature de la CE 5 Mars 2025

Date de début 1 Mars 2025

Date de fin 28 Février 2027

Financé au titre de

• The European Innovation Council (EIC)

Coût total

Aucune donnée

Contribution de l’UE

€ 2 499 990,63

Coordonné par

HQS QUANTUM SIMULATIONS GMBH
Germany

Periodic Reporting for period 1 - HQS-NextNMR (Next-Gen NMR Prediction and Analysis: A Cutting-Edge Tool for High-Precision Structure Identification)

Période du rapport: 2025-03-01 au 2026-02-28

Résumé du contexte et des objectifs généraux du projet

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.

Travail effectué depuis le début du projet jusqu’à la fin de la période considérée dans le rapport et principaux résultats atteints jusqu’à présent

During the current project phase, we established an end-to-end workflow for the prediction and analysis of NMR spectra, spanning quantum-chemical parameter generation, empirical parameter prediction, spectrum simulation, and quantitative comparison to experimental data.

We developed and deployed a scalable, automated quantum-chemical pipeline to compute high-quality NMR parameters using either local compute resources or HPC infrastructure. The workflow covers automated preparation, execution, monitoring, and post-processing of calculations. This enabled the population of an database containing quantum-chemically computed NMR parameters for more than 10,000 molecules, which is continuously extended as the project progresses.

Building on this foundation, we developed software to empirically predict the full set of NMR parameters based on the quantum-chemical database and conducted extensive benchmarks to quantify predictive performance. These two approaches address complementary technical needs: empirical prediction provides high speed and operational simplicity for rapid parameter generation. The quantum-chemical calculations set the basis for the empirical prediction with a consistent and systematically generated dataset.

In parallel to parameter prediction, we advanced spectrum computation by enhancing our solver and expanding its capabilities to include 2D NMR experiments. To enable robust comparison to experimental measurements and to prepare for automated analysis workflows, we developed methods to fit calculated spectra to experimental spectra and implemented quantitative measures to assess spectral similarity. These capabilities form the technical basis for mixture analysis, and corresponding methods and workflows are currently under active development within the running project.

To make the spectrum analysis capabilities conveniently accessible, we implemented a user-friendly web-based UI that integrates the underlying computational functionality and supports efficient, straightforward spectral analysis. This UI is already available as a free trial for potential customers and is being used to collect feedback relevant to further technical refinement during the ongoing project.

Finally, to assess the relevance and potential impact of quantum computing for predicting and analyzing NMR spectra, we prepared the technical groundwork for a strong benchmark calculation on real quantum hardware. This benchmark is planned for the next project phase and will enable a rigorous, hardware-based evaluation as the project continues.

Progrès au-delà de l’état des connaissances et impact potentiel prévu (y compris l’impact socio-économique et les conséquences sociétales plus larges du projet jusqu’à présent)

The analysis workflow for 1D NMR spectra that we have developed within the project is unique with regard to the capability of predicting and analyzing low-field NMR spectra. Furthermore the ab-initio calculation allows for a clear nuclei-to-peak assignment. That makes the analysis of NMR spectra faster but also allows for the development of (semi)-automatic analysis workflows. Furthermore, this is a huge opportunity for developing automated mixture analysis workflows. Several details in the different parts of the developed workflow are unique, e.g. the solver, the empirical prediction for the whole NMR parameter set, or the conformer sampling necessary to include all relevant structures in the quantum-chemical parameter generation. We have filed patent applications for the corresponding inventions.

The project will push the boundaries of NMR technology. By setting new standards for NMR prediction accuracy and low-field NMR utilization, the project provides new tools for molecular analysis relevant for research and for industry. To further develop the technology we need to bring the analysis workflow in lab environments. To reach the goal of automated mixture analysis we need further research and development efforts guided by examples and test cases with practical relevance.