Improved classification accuracy in 1- and 2-dimensional NMR metabolomics data using the variance stabilising generalised logarithm transformation

Abstract

Background: Classifying nuclear magnetic resonance (NMR) spectra is a crucial step in many metabolomics experiments. Since several multivariate classification techniques depend upon the variance of the data, it is important to first minimise any contribution from unwanted technical variance arising from sample preparation and analytical measurements, and thereby maximise any contribution from wanted biological variance between different classes. The generalised logarithm (glog) transform was developed to stabilise the variance in DNA microarray datasets, but has rarely been applied to metabolomics data. In particular, it has not been rigorously evaluated against other scaling techniques used in metabolomics, nor tested on all forms of NMR spectra including 1-dimensional (1D) 1 H, projections of 2D 1 H, 1 H J-resolved (pJRES), and intact 2D J-resolved (JRES).

Proteomic and metabolomic responses in hepatopancreas of Mytilus galloprovincialis challenged by Micrococcus luteus and Vibrio anguillarum

“…All one dimensional 1 H NMR spectra were converted to a data matrix using the custom-written ProMetab software in Matlab version 7.0 (the Mathworks, Natick, MA, USA) [38]. Each spectrum was segmented into bins with a width of 0.005 ppm between 0.2 and 10.0 ppm.…”

“…The total spectral area of the remaining bins was normalized to unity to facilitate the comparison between the spectra. All the NMR spectra were generalized log transformed with a transformation parameter λ = 1.0 × 10 − 8 to stabilize the variance across the spectral bins and to increase the weightings of the less intense peaks [38]. Data were mean-centered before principal component analysis (PCA) using PLS Toolbox (version 4.0, Eigenvector Research, Manson, WA).…”

Responses of Mytilus galloprovincialis to bacterial challenges by metabolomics and proteomics

“…All one dimensional 1 H NMR spectra were converted to a data matrix using the custom-written ProMetab software in Matlab version 7.0 (The MathsWorks, Natick, MA) [31]. Each spectrum was segmented into bins with a width of 0.005 ppm between 0.2 and 10.0 ppm.…”

Benzo(a)pyrene-induced metabolic responses in Manila clam Ruditapes philippinarum by proton nuclear magnetic resonance (1H NMR) based metabolomics

“…The area of each segment was calculated and normalized using the total integrated spectral area of the spectrum. All the NMR spectra were generalized log (glog) transformed (with transformation parameter, = 1 × 10 −8 ) to stabilize the variance across the spectral bins and to enhance the weightings of the less intense peaks (Purohit et al, 2004;Parsons et al, 2007). The data sets were preprocessed using mean-centering before principal components analysis (PCA) was performed using PLS Toolbox software (version 4.0,Eigenvector Research,Manson,WA).…”