Data Normalization for Diabetes II Metabonomics Analysis

Variable normalization alters the structure of data, thus affecting the outcome of multivariant analysis and calibration. The aim of this study was to evaluate the effect of two normalization methods, 1-norm and 2-norm, on the outcome of classification of ¹H nuclear magnetic resonance (NMR) metabonomic profiling, and to identify the characteristic metabolites that are responsible for the discrimination of plasma samples of diabetes II and healthy volunteers. The ¹H NMR spectra of human plasma from 14 healthy volunteers and 14 diabetes II volunteers were analyzed using principal component analysis (PCA) with the two normalization methods. The loading plots obtained by PCA using 1-norm and 2-norm were compared and same metabolites were identified that are responsible for the observed separations, with 2-norm method has better classification effect. These results demonstrated that NMR-based metabonomics analysis is capable of identifying the metabolites that are important for discriminating individuals of diabetes II from individuals of healthy.