Matrix Effects and Peak Shifts of NMR Spectra

One of the major challenges for automated data analyses and peak assignments for NMR spectra of biofluids from metabonomic studies is the effect of shifting peaks. As the matrix of biofluids and in particular urine is highly varying, the local environment of protons also varies. Thereby many parameters play a role such as changes of the pH, changes of the salt concentration, overall dilution of the sample, relative concentrations of specific ions, relative concentration of specific metabolites and many more. All these parameters can influence the shifts of peaks, whereby not all peaks are affected and different peaks are affected to a different extent even when belonging to the same metabolite. In the figure an example with ¹HNMR spectra of 30 samples from a human metabonomic study is shown. The clipping between 3.35 ppm and 3.49 ppm shows one singlet belonging to para-hydroxyphenylacetate and one triplet belonging to taurine. The position of the singlet is very stable at 3.455 ppm whereas the position of the taurine triplet varies between 3.407 ppm 3.435 ppm. As the pH of the samples was constant due to buffer added, the shifts of the taurine triplet are caused by different concentrations of salts, of metal ions, of metabolites, or by varying dilutions of the samples.

Clipping between 3.35 ppm and 3.49 ppm of 1H-NMR spectra of 30 samples from a human metabonomic study. It is visible that the location of the singlet of para-hydroxyphenylacetate at 3.455 ppm remains very stable for all samples whereas the location of the triplet of taurine varies between 3.407 ppm 3.435 ppm from sample to sample.

Although it has been demonstrated in a recent publication that peak shifts can be beneficial for a separation of different groups of samples under certain circumstances, this effect is unwanted for most applications. Two approaches have been proposed to handle variations in chemical shifts, whereby the first approach bases on a reduction of the spectral resolution by equidistant or non-equidistant binning methods, and the second approach bases on an alignment of peaks with the help of genetic algorithms, beam searches or sophisticated correlations. In this tutorial the widespread equidistant binning and the non-equidistant binning, which becomes increasingly popular, are shown.