COMPARATIVE STUDY OF DIFFERENT THIRD-ORDER DATA GENERATION APPROACHES

ALCARAZ, MIRTA R.; MONTEMURRO, MILAGROS; SIANO, GABRIEL; CULZONI, MARIA J.; GOICOECHEA, HÉCTOR C.

Congreso; XVI Chemometrics in Analytical Chemistry; 2016

p { margin-bottom: 0.25cm; line-height: 120%; }Over the last years,it has been demonstrated that the increase of multiway datadimensions has a positive impact on analytical figures of merit, e.g.higher sensitivity, lower limits of detection and quantitation,better selectivity, among others. First- and second-order dataanalyses have become excellent tools for the resolution of complexsamples which would result experimentally challenging from theunivariate calibration standpoint. On the other hand, even though noadditional analytical advantages have been yet proved, third-orderdata analysis for analytical applications constitutes a field worthto be explored [1]. Although multidimensional instrumental signalsare easy to be obtained with the available modern instrumentation,and several chemometric algorithms have been successfully developedto solve multi-way data problems, the way in which the multi-waydata are generated may have a significant effect on the finalresults. In this work, a comparative study of different third-orderdata generation approaches was carried out. Three methods based onidentical liquid chromatographic conditions but coupled to differentemission and excitation fluorescence detection systems were developedfor the quantitative analysis of antibiotics in aqueous matrices.The first approachincluded the collection of several fractions at the end of thechromatographic procedure by meansof a custom-built device that allows to collect fractions in a96-wells ELISA plate. Then, emission and excitation spectra wereregistered for every fraction by using a spectrofluorometer equippedwith a plate reader accessory coupled to an optical fiber and a gatedphotomultiplier. In this way, 25 emission-excitation matrices (EEM,size: 17×25) were obtained for each chromatographic run.[2] A secondstrategy was developed by using a 10 μL flow-cell connected at theend of the chromatographic instrument and placed in a fast-scanningspectrofluorometer. Here, it was possible to register sequential EEMsfor a unique chromatographic run. Since the fast-scannigspectrofluorometer takes only few seconds to register each EEM, itwas necessary neither to stop the chromatographic flow nor tocollect fractions after the chromatographic procedure, and 25sequential EEMs (size: 7×45) were obtained for each chromatographicrun. Finally, a multi-chromatographic run method involving a liquid chromatographic instrument coupled to a fast-scanning fluorescencedetector which allowed to register time-emission fluorescence datamatrices in a specific spectral range at a fixed excitationwavelength was developed. In order to build excitation-emission datamatrices, eight chromatographic runs at different excitationwavelengths (time-emission matrix size: 45×150) were required forthe same sample.The threemethodologies aforementioned were evaluated using differentalgorithms, such as PARAFAC, APARAFAC, PLS-RTL and MCR-ALS, andselectivity, sensitivity, robustness, and time processing wereevaluated. Since the data generation was different, each methodologyrequired a particular data pre-processing including smoothing, peakalignment, and baseline correction, among others. Furthermore, dueto differences in sensitivity provided by the implementation of avariety of detection mode it was necessary to assess several samplepreparation methods in order to reach good analytical figures ofmerit.