Databases and Software for NMR-Based Metabolomics

Abstract

New software and increasingly sophisticated NMR metabolite spectral databases are advancing the unique abilities of NMR spectroscopy to identify and quantify small molecules in solution for studies of metabolite biomarkers and metabolic flux. Public and commercial databases now contain experimental 1D ¹H, ¹³C and 2D ¹H-¹³C spectra and extracted spectral parameters for over a thousand compounds and theoretical data for thousands more. Public databases containing experimental NMR data from complex metabolic studies are emerging. These databases are providing information vital for the construction and testing of new computational algorithms for NMR-based chemometric and quantitative metabolomics studies. In this review we focus on database and software tools that support a quantitative NMR approach to the analysis of 1D and 2D NMR spectra of complex biological mixtures.

Keywords: Chemometrics; databases; nuclear magnetic resonance; quantitative metabolomics; software.

Fig. (1)

Overview of NMR metabolomics analysis workflow: The figure integrates the workflow involved in NMR metabolomics data analysis by both a chemometrics and “identify and quantify” approach. In a typical chemometrics approach, the workflow proceeds from data collection, processing, binning, and then multivariate analysis to identify spectral patterns of change. Metabolite identification, if performed at all, will occur after multivariate analysis. In an “identify and quantify” approach, the step of spectral identification occurs much earlier in the workflow, either after spectral processing or region of interest (ROI) segmentation. Model fitting is a technique that can be used in either workflow to create amplitudes to be used in the feature matrix. All of the features in the feature matrix (see Table 6) are actually derived values from the metabolite concentration.

Fig. (2)

L-alanine as an example of an interactive ¹H-NMR session in the MQMCD. The doublet at 1.47 ppm was selected in the NMR spectrum, and the corresponding methyl protons were highlighted in the Jmol structure (yellow atoms).

Fig. (3)

Spectrum from an E. coli extract (top) demonstrating use of regions of interest (ROIs) with rNMR [25]. Array of ROIs (red boxes) for multiple metabolites from a series of E. coli extracts (bottom) from rNMR.

Fig. (4)

Fast maximum likelihood reconstruction (FMLR) of a 2D ¹H-¹³C spectrum of liver extract performed by the Newton software package [19]. (Left) contour plot of a region of the ¹H-¹³C HSQC spectrum. (Middle) The FMLR reconstruction of the region. (Right) The corresponding residual. Annotations on the spectrum denote the centers of signals that were identified by FMLR. Signals from glucose (A) are much higher than those nearby from proline (B), and fructose (C,D,E). The volume of the observable residual in region A is less than 3% of the volume of the peaks. The amplitudes of the signals in the reconstruction are proportional to the concentration of the underlying species.

Fig. (5)

Overview of functionality and modules in the MetaboAnalyst 2.0 Package. Figure reproduced from [27].