PARTIAL LEAST-SQUARES REGRESSION OF 4TH-DERIVATIVE ULTRAVIOLET ABSORBENCY SPECTRA PREDICTS COMPOSITION OF PROTEIN MIXTURES - APPLICATION TO BOVINE CASEINS

A method is described for the quantitative determination of the composition of protein mixtures using fourth-derivative ultraviolet spectroscopy. The methodology was demonstrated by quantifying mixtures of the three main bovine caseins (alpha(s1)-, beta-, and kappa-casein). The 244-296-nm spectra region of protein mixtures of known composition, dissolved in 4 M guanidine hydrochloride, was used for the development of prediction models using the multivariate method of partial least-squares regression analysis (PLSR), implemented in a commercial software. The standard errors of prediction for 16 test samples were 13.4, 5.5, and 11.9% for alpha(s1)-, beta-, and kappa-casein, respectively, and the correlations between measured and predicted composition were 0.91, 0.99, and 0.94 for the three proteins, respectively. This method was found to be faster and simpler than alternate procedures (e.g., gel electrophoresis), suggesting the possibility of using this methodology as a quality control tool in industrial protein fractionation environments.