ImpuR: A Collection of Diagnostic Tools Developed in R in the Context of Peak Impurity Detection in HPLC-DAD but Potentially Useful with Other Types of Time-Intensity Matrices

• Main Authors: Christian Ritter, Jean Gilliard
• Format: Article
• Language: English
• Published: Foundation for Open Access Statistics 2007-01-01
• Series: Journal of Statistical Software
• Subjects: window-evolving factor analysis; spectral dissimilarity traces; spectroscopic
• Online Access: http://www.jstatsoft.org/v18/i09/paper

HPLC-DAD systems generate time intensity (absorbance) matrices called spectrochromatograms. Under good experimental conditions, spectro-chromatograms of elution peaks of pure analytes are bilinear products of a time peak and an absorbance spectrum. Co-eluting impurities create deviations from this pure bilinear structure. Unfortunately, other imperfections, such as scan averaging, large optical windows, imperfect lamp alignment, mobile phase fluctuations, etc. also create departures from the pure bilinear structure. This makes it hard to distinguish low concentration impurities from artifacts and hampers safe detection of contaminants. There are two main ways to deal with such artifacts: removal and simulation, and ImpuR provides R functions to do both and to integrate both approaches. More specifically, ImpuR provides a set of tools to explore time-intensity matrices with respect to their bilinear structure and departures from it. It includes exploratory graphs for bilinear matrices (bilinear residual graphs and singular value decompositions), spectral dissimilarity curves via window-evolving factor analysis with heteroscedasticity correction and the sine method, methods for removal of artifacts, and a comprehensive simulation tool to assess the impact of potential artifacts and to allow for the construction of guide curves for use with the sine method.