Chemometric Exploration of Quantitative NMR Data

TY - CHAP

T1 - Chemometric Exploration of Quantitative NMR Data

AU - Engelsen, Søren Balling

AU - Savorani, Francesco

AU - Rasmussen, Morten Arendt

PY - 2013

Y1 - 2013

N2 - This article outlines the synergistic relationship between NMR and chemometrics. The latent variable approach used in chemometrics has proven very powerful for performing inductive explorations of biological systems and for its usefulness insolving industrial problems effectively. This article reviews some of the commonest latent variable approaches applied to the exploratory and predictive modeling of NMR data. It describes how challenging NMR data can be adapted for multivariate data analysis and how the different chemometric methods manipulate the NMR data. The different results from unsupervised data exploration by principal component analysis and multivariate curve resolution are illustrated. On the other hand, many modern applications of NMR within metabolomics and quality control are based on supervised regression analysis or classification analysis. This article demonstrates how these basic chemometric methods work and gives examples of how such methods can be optimized by variable reduction and orthogonal factor extraction. Validation methods and classification performance by the receiver operating characteristics are illustrated. Finally, the potential for merging advanced multiway chemometric methods such as parallel factor analysis (PARAFAC) with the ability of NMR to record true high-order data is emphasized, and illustrated by the application to 2D diffusion-edited spectra of human plasma samples.

AB - This article outlines the synergistic relationship between NMR and chemometrics. The latent variable approach used in chemometrics has proven very powerful for performing inductive explorations of biological systems and for its usefulness insolving industrial problems effectively. This article reviews some of the commonest latent variable approaches applied to the exploratory and predictive modeling of NMR data. It describes how challenging NMR data can be adapted for multivariate data analysis and how the different chemometric methods manipulate the NMR data. The different results from unsupervised data exploration by principal component analysis and multivariate curve resolution are illustrated. On the other hand, many modern applications of NMR within metabolomics and quality control are based on supervised regression analysis or classification analysis. This article demonstrates how these basic chemometric methods work and gives examples of how such methods can be optimized by variable reduction and orthogonal factor extraction. Validation methods and classification performance by the receiver operating characteristics are illustrated. Finally, the potential for merging advanced multiway chemometric methods such as parallel factor analysis (PARAFAC) with the ability of NMR to record true high-order data is emphasized, and illustrated by the application to 2D diffusion-edited spectra of human plasma samples.

U2 - 10.1002/9780470034590.emrstm1304

DO - 10.1002/9780470034590.emrstm1304

M3 - Book chapter

VL - 2

SP - 267

EP - 278

BT - eMagRes

PB - Wiley

ER -