Classification of Condensed-Phase Infrared Spectra by Substructures Using Principal Components Analysis

Abstract

Expert system rules have been developed through the use of principal components analysis (PCA) for discriminating compounds containing large molecular substructures from their condensed-phase infrared spectra. With the use of this approach, the presence of substructures such as those of barbiturates, cocaines, and amphetamines can be recognized automatically. The classification rule can be generated from a PCA of a small training set of infrared spectra of compounds containing the substructure of interest. One important use of this type of expert system is the analysis of direct-deposition capillary gas and supercritical fluid chromatographic separations in which many peaks are eluted and analyzed by FT-IR spectrometry but only one or two contain the substructure of interest. Classification rules for substructures are easy to generate with little or no knowledge of characteristic group frequencies.