Third- and fourth-order data generation and modeling and their application in the analysis of complex systems

GOICOECHEA H C

Congreso; XX Euroanalysis; 2019

Over the last years, several analytical applications based onmultiway calibration have been developed aiming to improvethe performance of the methodologies by exploiting the benefits offered by the higher-order data modeling. In this regard, itis noticeable that third- and fourth-order data-based analyticalmethods are still scarce despite all the aids that have been alreadydemonstrated [1]. For instance, a) the possibility of decomposinga unique data array for each sample; b) the enhancement of theanalytical figures of merit and c) the possibility to tackle collinearity problems arisen from the introduction of an extra mode, areincluded among the benefits that have been proved and representthe so-called ?third-order advantage?. This topic has been a matterof debate among the scientific community and has promoted intensive research towards unravelling its existence and true nature.In this work, different applications focused on third-orderexcitation-emission fluorescence matrices (EEM)-based data aredescribed. Instrumental configurations, data generation and datamodeling are depicted.First, a chromatographic procedure coupled to on-line EEMacquisition was developed for the analysis of emergent contaminant in aqueous environmental samples. With this approach, itwas possible to acquire 25 sequential EEM for a given chromatographic run. Neither chromatographic flow stop nor fraction collection after chromatographic procedure were required. However,to the best of our knowledge, there is no chemometric algorithmcapable to model this kind of data by virtue of the strong dependence between instrumental modes. Thus, an algorithm basedon Tucker 3 decomposition is proposed as alternative to modelnon-trilinear/quadrilinear 4-way data.In order to provide evidences of the real existence of thethird-order advantage, a system consisting on the determination of 3 analytes with identical fluorescence spectra by means ofthe photodegradation kinetics was developed. For this purpose,the photodegradation of the analytes was monitored by EEM indifferent UV-irradiation periods of time. Then, the three possible second-order calibration models were compared against thethird-order calibration model in terms of predictive capability,figures of merit and retrieved spectral profile evaluation.Last, study of the evolution of a cell culture registering EEMs.A sterile sample from standard etanercept (Et) process was acquired daily for viable cells count and determination of Et concentration in supernatant by standard off-line univariate techniques.EEM were simultaneously obtained (second-order data) and thenarranged in a third-order structure. The later was alternatively modeled by different chemometric algorithms. Firstly, unsuperviseddecomposition methods Parallel Factor Analysis (PARAFAC)and Multivariate Curve Resolution (MCR) were considered toqualitatively analyse the spectral information. PARAFAC components were putatively related to biological fluorophores presentin the culture media during fermentation of CHO cells (aromaticaminoacids, pyridoxine, flavin, folic acid and NAD) demonstrating that a proper modeling of a three-way array corresponding toa single fermentation process, allows extracting extremely usefulinformation about it.Reference:[1.] G.M. Escandar, H.C. Goicoechea, A. Muñoz de la Peña, A.C. Olivieri, Second- and higher-order data generation and calibration: Atutorial. Anal. Chim. Acta 806 (2014) 8? 26.