Multivariate analyses of near-infrared and UV spectral data.
Access changed 8/1/11
Various chemometric analyses were applied to spectroscopic data with goals to develop alternative methods that could be employed in government or industrial settings. With the concerns of these organizations in mind, the described methods are cost-effective and time-efficient. The first method is aimed at establishing a time of death from skeletal remains—an issue that continues to be difficult for the forensic community. Following death, the skeletal remains undergo changes in chemical composition. This includes the breakdown of protein and the loss of water. Near-infrared spectroscopy is sensitive to vibrations associated with both protein and water. In the described method, near-infrared reflectance measurements of aging porcine skeletal remains were correlated to postmortem interval (PMI). Initial studies were conducted to determine the optimum sampling orientation—cross-sectional or surface. Several chemometric approaches were investigated, but the best results were obtained through a scheme involving classification by partial least-squares discriminant analysis (PLS–DA) followed by segmented partial least-squares regression (PLSR). The method was evaluated through independent test sets. The optimized method was able to predict PMI with an average deviation of six days. A brief field study was also conducted and yielded similar results. The second study relates to a present analytical encumbrance faced by the pharmaceutical industry, namely assuring the enantiomeric purity of chiral active pharmaceutical ingredients (APIs). With the rising number of chiral drugs on the market, the analytical burden continues to increase. Ultraviolet absorption spectral data were correlated to enantiomeric composition by PLSR for solutions containing a chiral analyte and a chiral ionic liquid (IL) as a chiral selector. Test set evaluation gave results of average deviations of ± 4.0–12 units of %D depending on the analyte and chiral IL involved. Finally, a quality control analysis was demonstrated, which follows a classification format where the sample either meets or does not meet the specified requirement regarding enantiomeric purity. Test set evaluation gave results of 97% correct classifications for a threshold of 1% impurity.