Quantitation of resonances in biological 31P NMR spectra via principal component analysis: potential and limitations.

This paper examines the potential and limitations of peak area quantitation of biological NMR spectra using principal component analysis (PCA), including its requirement for prior knowledge. The principles of the method are presented without in-depth mathematical treatment. PCA is illustrated for simulated data, 31P NMR spectra obtained consecutively over 1-2.5 days from perfused Rat-2 cells metabolizing the choline analogue phosphoniumcholine (Chop) and in vivo proton-decoupled, NOE-enhanced, three-dimensional CSI localized 31P NMR spectra of the liver of healthy volunteers. The results show that PCA can be used to quantitate strongly overlapping peaks without prior knowledge of the peak shapes or positions and to reconstruct spectra with significantly reduced noise variance. Two major limitations of PCA are presented: (1) PCA cannot separate peaks whose intensities are well correlated; (2) PCA is sensitive to differences in chemical shift and line-width of peaks between spectra. The discussion focuses on what knowledge of the biological and spectroscopic features of the samples and the principles of PCA is necessary for peak area quantitation via PCA.