Work

Automated spin system analysis in NMR spectroscopy with SpinDETR : a deep learning approach

Year: 2024

Type: conference-paper

Abstract: lD nuclear magnetic resonance (NMR) spectroscopy is pivotal in small molecule analysis across chemistry, biochemistry, and pharmaceutical domains. However, interpreting complex spectra often requires specialized expertise and time. Existing automated methods often perform well on simple spectra but usually rely on prior molecular knowledge and struggle with complex spin systems. Transformer-based deep learning models, previously successful in natural language processing and computer vision, are emerging as potential game-changers for NMR spectral analysis. Our automated method, SpinDETR, revolutionizes NMR data analysis, particularly for small molecules with complex spin systems featuring strong coupling and overlapping signals. This innovative approach accurately predicts crucial spin system parameters, such as chemical shifts and coupling constants, directly from NMR spectra without prior molecular knowledge. SpinDETR significantly accelerates the analysis process, broadens the scope of applications, and unlocks new opportunities in analytical chemistry and drug discovery, making it a transformative method in NMR spectroscopy.

Source: Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences)

DOI: 10.21256/zhaw-31443

Authors: Schmid, Nicolas, Wanner, Marc, Fischetti, Giulia, Meshkian, Mohsen, Bruderer, Simon, Henrici, Andreas, Wegner, Jan Dirk, Sigel, Roland K.O., Heitmann, Björn, Wilhelm, Dirk

Language: English

Topic: Advanced NMR Techniques and Applications

Subfield: Spectroscopy

Field: Chemistry

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W6888566180

FWCI:

Cites:

Cited by:

Related to: