Univariate Statistical Analysis as a Guide to ¹H-NMR Spectra Signal Assignment by Visual Inspection.

Chenglin Zhu,Beatrice Vitali,Gilbert Donders,Luca Laghi,Yan Li,Carola Parolin

doi:10.3390/metabo9010015

Chenglin Zhu, Beatrice Vitali + Show 4 more

Open Access

https://doi.org/10.3390/metabo9010015

Copy DOI

Journal: Metabolites	Publication Date: Jan 15, 2019
Citations: 10	License type: CC BY 4.0

Affiliation: University of Bologna, University of Antwerp

Abstract

In Proton Nuclear Magnetic Resonance (1H-NMR) spectroscopy, the signals assignment procedure is normally conducted by visual inspection of the spectra, by taking advantage of the innate predisposition of human eye for pattern recognition. In the case of untargeted metabolomics investigations on food and body fluids, the complexity of the spectra may lead the user to overlook signals, independently from their biological relevance. Here, we describe a four steps procedure that is designed to guide signals assignment task by visual inspection. The procedure can be employed whenever an experimental plan allows for the application of a univariate statistical analysis on a point-by-point basis, which is commonly the case. By comparing, as a proof of concept, 1H-NMR spectra of vaginal fluids of healthy and bacterial vaginosis (BV) affected women, we show that the procedure is also readily usable by non-experts in three particularly challenging cases: overlapping multiplets, poorly aligned signals, and signals with very poor signal-to-noise ratio. The paper is accompanied by the necessary codes and examples written in R computational language to allow the interested user gaining a hands-on impression of the procedure’s strengths and weaknesses.

Highlights

1 H-NMR spectroscopy represents one of the election techniques for investigating the metabolome of food and body fluids, defined as the ensemble of their small metabolites that were observed comprehensively [1]
Researchers try to make their observations on the metabolome really comprehensive by treating 1 H-NMR spectra to fingerprints of the samples under investigation [2,3,4]
Based on criteria that are described in greater detail elsewhere [12], 43 women affected by bacterial vaginosis (BV) and 37 age-matched healthy women (H) were recruited

Summary

Introduction

1 H-NMR spectroscopy represents one of the election techniques for investigating the metabolome of food and body fluids, defined as the ensemble of their small metabolites that were observed comprehensively [1]. Researchers try to make their observations on the metabolome really comprehensive by treating 1 H-NMR spectra to fingerprints of the samples under investigation [2,3,4]. They do so by considering the points constituting each spectrum as variables by means of multivariate analysis, with no attempt to ascribe the NMR signals to specific molecules. The task of signals assignment has been faced by the first works dealing with metabolomics by means of two-dimensional (2D) experiments [6]. As a step forward, the creation of public databases, listing chemical shift and multiplicity of the signals of pure compounds, has allowed the high throughput identification

Methods

Discussion

Conclusion