Determination of human serum albumin and γ-globulin in a control serum solution by near-infrared spectroscopy and partial least squares regression

Near infrared (NIR) spectra in the 1300– 1850 nm region were measured for control serum solutions containing both albumin and γ-globulin of various concentrations. Partial least squares two (PLS2) regression was applied to the NIR spectra to determine simultaneously the concentrations of both proteins. For albumin, the correlation coefficient (R) of 0.988, the standard error of calibration (SEC) of 1.61 g/L, the standard error of prediction (SEP) of 1.29 g/L, the relative standard deviation (RSD) of 0.026 and the ratio of standard deviation of reference data in prediction to SEP (RPD) of 12.2 were obtained. For γ-globulin, the corresponding values were 0.997, 1.36 g/L, 1.35 g/L, 0.0365 and 8.66, respectively. The regression coefficients (RCs) of PLS factors were compared between albumin and γ-globulin, and the observed differences in the RCs were discussed based upon the differences in the hydration between albumin and γ-globulin. In order to explore the effects of various metabolites such as glucose, and cholesterol on the chemometrics models, the RCs for albumin and γ-globulin in the control serum solutions were also compared with those for albumin and γ-globulin in phosphate buffer solutions previously studied. The results of our experiments show that NIR spectroscopy with the use of PLS2 regression has considerable promise in nondestructive determination of the concentrations of blood serum proteins. Received: 31 December 1997 / Revised: 9 April 1998 / Accepted: 27 April 1998