Comparison of computational approaches for identification and quantification of urinary metabolites in 1H NMR spectra. Issue 18 (25th April 2018)

• Record Type: Journal Article
• Authors: Cassiède, Marc; Mercier, Pascal; Shipley, Paul R.; Dueck, Meghan; Kamravaei, Samineh; Nair, Sindhu; Mino, James; Pei, Lei; Broadhurst, David; Lacy, Paige; Quémerais, Bernadette
• Is Part Of: Analytical methods. Volume 10:Issue 18(2018)
• Page Start: 2129
• Page End: 2137
• Publication Date: 2018-04-25
• Subjects: Chemistry, Analytic -- Periodicals; Analytical biochemistry -- Periodicals; Chemical laboratories -- Standards -- Periodicals; 543.1905
• DOI: 10.1039/c8ay00830b
• Languages: English
• ISSNs: 1759-9660

Abstract

Abstract : A Monte Carlo simulation technique is used to accurately measure metabolite concentrations in urine. Abstract : Nuclear magnetic resonance (NMR) spectroscopy is extensively used in analytical chemistry as a powerful, non-invasive, and non-destructive tool to elucidate detailed structures of small molecules in complex mixtures. A major initiative in NMR is the identification of metabolic changes in biological fluids, particularly urine, as potential biomarkers for specific diseases or occupational exposure. However, major challenges are encountered during data processing of complex NMR spectra, presenting obstacles in the use of NMR analysis in clinical applications. In this report, metabolite concentrations were determined using three different computational approaches with complex NMR spectra obtained using 33 replicates of quality control (QC) human urine samples. We have used a new computational method involving Monte Carlo (MC) simulation to automatically deconvolve and quantify metabolites in NMR spectra from human urine. MC simulation is independent of experimental bias or human error, and is recommended as the least biased approach to peak fitting for NMR spectra derived from human urine samples. We found that similar results could be obtained using MC simulation in urine samples compared with two previous approaches that are subject to experimental bias and/or human error.