Metabolomics has been widely applied in human diseases, nutrition, and environmental science areas, and it studies the systematic changes of metabolites over various types of stimuli. The NMR-based metabolomics has the advantages of high reliability, nondestructive and limited sample processing but suffers from peak overlap problems which limited the accuracy of quantitative analysis. Two-dimensional (2D) NMR has an advantage in solving the one-dimensional (1D) NMR overlap problem, but the data interpretation of 2D NMR especially for the quantitative analysis is difficult. Our recent study has built an approach to isolate the potential peaks from the 2D NMR, however, the peak selection accuracy is still changeling. In this study, we applied machine learning approaches including random forest and artificial neuron network (ANN) to 2D NMR TOCSY spectra for various types of samples including blueberry extracts, sea anemones extracts, tea extracts, and milk samples. Our results show that ANN has high accuracy in the selection of 2D NMR peaks. Our study provides a pioneering study to build an automatically processing 2D NMR in metabolomics study which pave the road for 2D NMR applications in regular metabolomics studies.

Evaluation of machine learning approaches in two dimensional NMR data interpretation for metabolomics studies

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.