Reconstruction of HMBC Correlation Networks: A Novel NMR-Based Contribution to Metabolite Mixture Analysis

Abstract

A new in silico method is introduced for the dereplication of natural metabolite mixtures based on HMBC and HSQC spectra that inform about short-range and long-range H-C correlations occurring in the carbon skeleton of individual chemical entities. Starting from the HMBC spectrum of a metabolite mixture, an algorithm was developed in order to recover individualized HMBC footprints of the mixture constituents. The collected H-C correlations are represented by a network of NMR peaks connected to each other when sharing either a ¹H or ¹³C chemical shift value. The network obtained is then divided into clusters using a community detection algorithm, and finally each cluster is tentatively assigned to a molecular structure by means of a NMR chemical shift database containing the theoretical HMBC and HSQC correlation data of a range of natural metabolites. The proof of principle of this method is demonstrated on a model mixture of 3 known natural compounds and then on a real-life bark extract obtained from the common spruce (Picea abies L.).