| Abstract: | An algorithm of decomposition of protein tryptophan spectra into components was developed. The spectral shape of components is described by a uniparametric log-normal function. Rise of certainty and accuracy of resolution of widely overlapping smooth spectral components (a typical uncorrect reverse problem) was achieved using several regularizing factors: (i) the set of experimental spectra used were measured at several quencher concentrations; (ii) the functional being minimized, along with the root mean square residuals of intensities, the term depending on the obedience to the Stern-Volmer law; (iii) an extra information is used--the number of experimental values greatly exceeds the number of parameters to be estimated. The minimum of functional is determined by a consecutive setting of all possible combinations of component spectral maxima values, which allows to avoid sticking in the local minima of noisy functional. The real experimental noise restricts the decomposition into not more than three components. The decomposition error does not exceed the experimental one. The algorithm functioning is illustrated by resolution of tryptophan fluorescence spectra of papain into one, two, and three components. |
