Visualizing metabolic changes in breast-cancer tissue using 1H-NMR spectroscopy and self-organizing maps.

In-vitro NMR spectroscopic examinations of tissue extracts can be combined with appropriate pattern-recognition and visualization techniques in order to monitor characteristic metabolic differences between tissue classes. In the present study, such techniques are applied to a set of 88 breast-tissue samples with the intention of identifying typical differences between various tissue classes. The set contains 49 breast-tumor samples of various tumor grades and 39 samples of healthy tissue. The metabolite compositions of the tissue extracts were investigated using a dual extraction technique and high-resolution (1)H-NMR spectroscopy. The spectra of the hydrophilic and the lipophilic compounds were assigned to three groups according to different malignancy grades of the respective tissue samples. The group characteristics were analyzed using the k-nearest-neighbor method and self-organizing-map visualizations. The results show an increase of UDP-hexose, phosphocholine and phosphoethanolamine concentrations according to the tumor grade. Higher concentrations of taurine were detected in the malignant samples. Myo-inositol and glucose content were elevated in control samples compared with malignant tissue. Both compounds also characterized different subgroups in the pool of unaffected tissue samples depending upon fat content or fibrosis. Several lipid metabolites showed a characteristic elevation with high malignancy. Copyright 2003 John Wiley & Sons, Ltd.