Boosting annotation in nutrimetabolomics by Feature-Based Molecular Networking: analytical and computational strategies applied to human urine samples from an untargeted LC-MS/MS based bilberry-blueberry intervention study

Abstract

Urine represents a challenging metabolite mixture to decipher. Yet, it contains valuable information on dietary intake patterns as typically investigated using randomized, single-blinded, intervention studies. This research demonstrates how the use of Feature-Based Molecular Networking in combination with public spectral libraries, further expanded with an 'In-house' library of metabolite spectra, improved the non-trivial annotation of metabolites occurring in human urine samples following bilberry and blueberry intake. Following this approach, 65 berry-related and human endogenous metabolites were annotated, increasing the annotation coverage by 72% compared to conventional annotation approaches. Furthermore, the structures of 15 additional metabolites were hypothesized by spectral analysis. Then, by leveraging the MzMine quantitative information, several molecular families of phase II (e.g., glucuronidated phenolics) and phase I (e.g., phenylpropionic acid and hydroxybenzoic acid molecular scaffolds) metabolism were identified by correlation analysis of postprandial kinetics, and the dietary impact of endogenous and exogenous metabolites following bilberry-blueberry intake was estimated.

Rationalising the retro‐Diels‐Alder fragmentation pattern of viscutins using electrospray interface‐tandem mass spectrometry coupled to theoretical modelling

“…Therefore, employing several tools in conjunction increases the accuracy in the identification and characterisation of natural products. As such, computational tools such as molecular networking 13–15 and MS2LDA 16,17 have been developed, and they allow the visualisation of compounds that share unique fragmentation patterns and substructures as molecular families. The introduction of molecular networking and its complementary tools has enabled mass spectral mining from multiple data sets 18,19 .…”