PRINCIPAL COMPONENT ANALYSIS FOR ARTIFACT REDUCTION IN COSY SPECTRA

The use of principal component analysis for removing artifacts from baseline offsets, baseline roll, t, noise, and spurious signals such as quadrature images from COSY 2D NMR spectra is described. The method works by deriving and subtracting major components common to the largest number of points in the data set. In a typical COSY spectrum, the major components, which account for more than 95% of the variance, are t1 and t2 ridges which arise from baseline offset, baseline roll, t1 noise, quadrature images, and diagonal signal intensity. In all of the data sets examined, PCA showed the ability to progressively remove artifact signals first. If components accounting for more than 95-99% of the variance are removed, real signals are subtracted from the data set; however, creation of peaks which were not already present in the spectrum is not observed. © 1990.