Quantitative Analysis in the Presence of Spectral Interferents Using Second-Order Nonbilinear Data

One of the most serious problems that can occur when a model is being used to predict the concentration of an analyte In an unknown sample is the presence of one or more chemical species that are unaccounted for in the calibration samples. With very few exceptions, the model being used to predict analyte concentration is invalidated by the presence of these spectral interferents, and in some cases It is not even possible to detect the invalidation of the model except when nonsensical predictions are obtained. One group of methods that have been used successfully for prediction in the presence of spectral interferents are the rank annihilation methods. This paper compares nonbilinear rank annihilation with three curve resolution methods on three data sets (simulated spectra, two-dimensional nuclear magnetic resonance, and tandem mass spectrometry) for their abilities to accurately predict the concentration of an analyte in the presence of one or more spectral Interferents. Multiple linear regression is used as a referee method. It is shown that nonbilinear rank annihilation is the only one of the methods tested which has any potential for solving real chemical problems, lnasmuch as the curve resolution methods are not consistently able to correctly solve even idealized cases (with no noise present).