Multivariate calibration models based on the direct analysis of near-infrared single-beam spectra

Abstract

Multivariate calibration models are generated for glucose, glutamine, and asparagine on the basis of partial least-squares regression analysis of near-infrared single-beam spectra covering the 5000-4000-cm-1 (2000-2500-nm) spectral range. Models are constructed with both raw and digitally Fourier-filtered single-beam spectra. Model performance is evaluated and compared with that of analogous models constructed from the corresponding computed absorbance spectra. Five unique data sets are examined corresponding to the measurement of (A) glucose in phosphate buffer with different temperatures, (B) glucose with variable albumin protein levels, (C) glucose with variable triacetin levels, (D) glucose and glutamine in a set of binary mixtures, and (E) glutamine and asparagine in a set of binary mixtures. In all cases, models based on single-beam spectra perform as well as those based on computed absorbance spectra.