REMOVAL OF SPECTRAL NOISE IN THE QUANTITATION OF PROTEIN-STRUCTURE THROUGH INFRARED BAND DECOMPOSITION

The underlying noise in the infrared spectra of proteins may introduce artifacts in the quantitation of proteins by curve-fitting of the ami de I band. Smoothing methods are able to reduce the noise but can intr oduce alterations in band shape that affect the information contained in the spectrum. Three methods to remove noise-Savitzky-Golay, Fourier filtering, and maximum entropy-have been used to ascertain their infl uence on the quantitative information when applied to protein bands. U se of artificial curves shows that whereas Savitzky-Golay and Fourier smoothing are able to reduce the noise, they distort the band shape. M aximum entropy is more efficient in reducing the noise in artificial c urves with added noise, and provided a narrowest bandwidth below 12 cm (-1), no bandshape distortion is obtained. Using the smoothing in natu ral spectra, the presence of spurious bands in the initial parameters coming from artifacts introduced by deconvolution or derivation is red uced. Moreover, the dispersion in the percentage area values in a seri es of similar spectra is also decreased below 2%, a value that discrim inates the effect of ligand binding to proteins. The maximum entropy m ethod is proposed as a tool to improve the quantification of protein s tructure by infrared spectroscopy. (C) 1997 John Wiley & Sons, Inc.