The Need

Like a molecular “fingerprint”, NMR spectra provide valuable insights into the properties and structures of molecular compounds. However, analyzing and interpreting NMR spectra is inherently challenging as comprehensive and explicit identification of a compound is tedious and time-consuming. Moreover, samples being analyzed often contain multiple compounds, creating spectra overlap, or “overlapping fingerprints”, further complicating NMR spectra analysis and making analysis difficult even for highly trained specialists. New software that can allow the broader scientific community to rapidly analyze NMR spectra would significantly enhance its applicability in many fields.

The Technology

A suite of software programs called “DEEP Picker” and “Voigt Fitter”, they use a deep neural network for automating the analysis of 1-D and 2-D NMR spectra. They use seven convolutional layers and are trained on an extensive database of spectra for known compounds. The inventors have created a software prototype and showed it accurately identifies multiple proteins and small molecules, such as metabolites, in complex mixtures, despite the substantial spectra overlap.

Commercial Applications

This invention can be used for drug development, basic research, quality control, water treatment, food processing facilities, etc.

Benefits/Advantages

Compared to existing NMR products, this technology facilitates rapid and improved analysis of NMR spectra, when necessary, with full quantitation, thereby accelerating research and quality for the evaluation of complex materials.

Patent Pending

WO2024049537A1