Investigations into Background Correction and Retention Time Alignment to Enhance Quantitative Chemometric Analysis of Comprehensive Two-Dimensional Liquid Chromatography-Diode Array Detection Data

摘要

The focus of the projects presented here was to develop possible solutions to three issues commonly encountered during chemometric analysis of comprehensive two-dimensional liquid chromatography diode array detector (LCxLC-DAD) data. The focus of the first project was to determine a means of performing background correction that removed two background ridges. The methods of simply subtracting out a mean blank sample, singular value decomposition based background correction (SVD-BC) and asymmetrically weighted least squares (AWLS) were compared. AWLS was found to be the only background correction technique to fully remove the ridges. However, AWLS was also found to attenuate the peak intensity by approximately 25% due to over fitting of the background at the lower wavelengths. The focus of the second project was the investigation of five common interpolation strategies for the reconstruction of the sampled first dimension peak. The interpolation strategy that best reproduced the original first dimension retention time was Gaussian fitting. This was expected given that the simulated data set was generated using a Gaussian model for the peak shape. An algorithm, semi-automated alignment method (SAAM), was then developed that allowed for each peak to be aligned independently of the other peaks in the data set. SAAM was validated using both simulated and experimental data. The simulated results indicated that SAAM produced percent recoveries close to 100%. SAAM was also compared to iterative key set factor analysis-alternating least squares (IKSFA-ALS) for the analysis of phenytoin in a waste water treatment plant effluent. SAAM produced a concentration of 26±3 ppb compared to 39±9 ppb from IKSFA-ALS. While these results are very different, the result produced by SAAM is still within the experimental error of the reference 2D-LC/MS/MS method, 42±19. Finally, SAAM was compared to two existing literature methods. A mixture of simulated and experimental data sets was used to measure the accuracy and precision of the results. SAAM was found to be impacted less by intra- and inter-sample retention time shifting then PARAFAC2. SAAM and shifted candecomp/PARAFAC were found to produce very similar results. However, SAAM was found to experience some difficulty producing accurate and precise results with some of the experimental data sets.