Four-order chromatographic-fluorescence data arrays modeling for the quantitation of fluoroqunolones in environmental samples

GOICOECHEA, H; CULZONI M J; ALCARAZ

Philadelphia

Congreso; Pittcon; 2013

Owing to the prevalent acceptance in the analytical community about the need to study increasingly complex samples by improving analytical methods, an intensive researching of new multi-way data generation and its consequent modeling with convenient algorithms have been carried out in lasts years [1]. In this context, four-order chromatographic data arrays modelled with appropriate algorithms appear as an interesting tool for solving complex analytical tasks. In most of the literature applications, the well known parallel factor analysis (PARAFAC) algorithm has been the choice for processing four-way arrays [2]. Moreover, a methodology based on the use of latent variables was adapted for processing four-way data, maintaining the second-order advantage: the unfolded variant of PLS (U-PLS) coupled to RTL, which calls for extensive studies [3]. Fluoroquinolones are broad-spectrum antibiotics that play an important role in treatment of serious bacterial infections, especially hospital-acquired infections and others in which resistance to older antibacterial classes is suspected. Third-order data were generated through a double chromatographic running of the sample and recording excitation and emission spectra, respectively. A previous selection of the most convenient wavelengths was performed in each case. Then, hypermatrices time-excitation-emission were arranged for each sample. Four-way arrays (sample plus standards data) were modeled with PARAFAC and U-PLS/RTL, after a convenient data preprocessing (shift and baseline correction). Validation samples (mixture solutions prepared in milli-Q water) and real test samples (tap water and well water) were analyzed. As a conclusion, five fluoroquinolones could be quantitated at pg/L levels with acceptable figures of merit as indicator of accuracy (mean recoveries:?) and precision (CV%=) were obtained.