Metabolic phenotyping of human plasma by 1H‐NMR at high and medium magnetic field strengths: a case study for lung cancer

• Published in: Magnetic resonance in chemistry 2017-08, Vol.55 (8), p.706-713
• Main Authors: Louis, Evelyne, Cantrelle, Francois‐Xavier, Mesotten, Liesbet, Reekmans, Gunter, Bervoets, Liene, Vanhove, Karolien, Thomeer, Michiel, Lippens, Guy, Adriaensens, Peter
• Format: Article
• Language: English
• Subjects: 1H, magnetic field strength; Biochemistry; Blood plasma; Cancer; Classification; Field strength; Lung cancer; Magnetic fields; metabolic phenotype; Metabolites; NMR; Nuclear magnetic resonance; nuclear magnetic resonance spectroscopy; Patients; Plasma; Signal to noise ratio; Spectral resolution; Spectrometers; Spiking
• ISSN: 0749-1581; 1097-458X
• DOI: 10.1002/mrc.4577

Accurate identification and quantification of human plasma metabolites can be challenging in crowded regions of the NMR spectrum with severe signal overlap. Therefore, this study describes metabolite spiking experiments on the basis of which the NMR spectrum can be rationally segmented into well‐defined integration regions, and this for spectrometers having magnetic field strengths corresponding to 1H resonance frequencies of 400 MHz and 900 MHz. Subsequently, the integration data of a case–control dataset of 69 lung cancer patients and 74 controls were used to train a multivariate statistical classification model for both field strengths. In this way, the advantages/disadvantages of high versus medium magnetic field strength were evaluated. The discriminative power obtained from the data collected at the two magnetic field strengths is rather similar, i.e. a sensitivity and specificity of respectively 90 and 97% for the 400 MHz data versus 88 and 96% for the 900 MHz data. This shows that a medium‐field NMR spectrometer (400–600 MHz) is already sufficient to perform clinical metabolomics. However, the improved spectral resolution (reduced signal overlap) and signal‐to‐noise ratio of 900 MHz spectra yield more integration regions that represent a single metabolite. This will simplify the unraveling and understanding of the related, disease disturbed, biochemical pathways. Copyright © 2017 John Wiley & Sons, Ltd. This study describes metabolite spiking experiments on the basis of which the 1H‐NMR spectrum can be segmented into well‐defined integration regions, and this for 400 and 900 MHz spectra. Subsequently, the integration data of a case–control dataset were used to train a multivariate classifier for both magnetic field strengths. The discriminative power of the resulting models is rather similar, i.e. a sensitivity and specificity of 90 and 97% for the 400 MHz data versus 88 and 96% for the 900 MHz data.