8290614321

7. Opublikowane badania własne

l Orzeł M Dauykowild j Chrmowma cnd InieHycftii Labonuory Systems 137 (2014) 74 At

all of the compounds of interest. Therefore. in order to obrain informa-tion about a wide rangę of fluorophores. a number of emission spectra are collected by the sequential exritation of a sample at certain wavelengths. These emission spectra are then arranged into a so-called excitarion~emission fluorescence matrix (EEM). which repre-sents a 2-0 analytical signal containing emission spectra collected for parTicular excitation wavelengths in rows. Considering the relatively high information content. an EEM can be treated as a sample fingerprint that potentially resembles information about its TAC value. For instance, EEMs were used as fluorescence fingerprints of diesel oil samples in order to determine the concentrations of excise tax components [15] or to detect rhe illegal process of fuel laundering (the removal of excise duty components) 116). They have also been considered to be useful analytical signals to monitor the quality of food products such as sugar 117]. yoghurt (13) and edible oils ] 18|. Bearing in mind the advantages of fluorescence fingerprints. a novel approach to evaJuate the TAC of food samples was developed based on their EEMs. Multivariate calibra-bon methods were applied in order to extract useful Chemical information from EEMs that described the TAC value. The TAC values were obtained from rwo reference analytical assays - the ORAĆ assay |7] and the total phenol content of samples assay. TPC119]. To illustrate the performance of the proposed method, rwo food commodiries such as coflee and peppermint. both of which have been relatively well described in the literaturę, were selected. ORAĆ and TPC values for each sample were expressed as the Trolox (TE) and gallic acid (CA) equivalents. respectively. The EEMs were registered in selected excita-tion and emission ranges. Then. the collected fluorescence fingerprints were related to their corresponding response TAC values using the par-oal least squares regression (20). PLS and the N-way partial least squares |21 J. N-PLS. The proposed method is characterized by its figures of merit (limits of detection and quantitarion) as calculated for the optima! calibration models.

2. Materials and methods

2.1.    Theory

2.1.1.    Dererminańon of rhe total phenohc conrerrr

The total polyphenol content of samples was determined using the Folin-Ciocalteau method as described in detail in reference [19|. The calibration curve. which served as the standard, was constmcted for gallic arid and the absorbance was recorded at 760 nm.

2.12. The oxygrn rudical absorbance capadty assay

The response of the ORAĆ assay relies on the damage to a fluorescent agent - fluorcscein, which is caused by free radicals and is monitored as the decrease of the fluorescence intensity over a certain period of time. The antioxidants that are present in a sample inhibit damage to the fluorescent agent and thus the fluorescence signal is sustained. Simulta-neously. the fluorescence intensity of a blank probe (without antioxi-dants, but with the same amounts of a free radical generator and a fluorescence agent as in the original sample) is examined. These two Chemical reactions are drwen to completion. The TAC is expnessed as the area hetween the two curves that described the decrease of fluorescence intensity over dme for the blank and tested sample. respecbvely. In our study, a calibration curve was built for a set of standards with known TAC values. The water solu We vitamin E analog, which is called Trolox (6-hydroxy-2,5.7,8-tetramethylchroman-2-carboxylic acid). served as the standard antioxidant for the ORAĆ assay.

2.1.3. Data modeling

Excitation-emission fluorescence matrices contain multiple fluorescence measurements, which result in 2-D spectra. For a set of samples. a collection of EEMs can be organized as a 3-D data array |22| with the dimensions exci (arion wavelength x emission wavelengrhs x samples. To extract the relevant Chemical information from the EEMs. advanced 75

chemometric preprocessing and modeling methods are necessary. One of the most important preprocessing steps for the EEMs consiscs of the detection and elimination of the Rayleigh and Raman scattering. The scarrering is manifested on the EEMs as diagonal lines of peaks (see Fig. la). It is a chemicalły irrelevant component and can significandy influence the shapc of the characterisbc of the peaks for key fluorophores. Therefore. prior to the construction of the model, the scattering musr be corrected. In this study. correcbon of scattering was done using the approach described earlier (23|, alrhough there are other methods also performing well |24). Peaks corresponding to scattering were deteaed. removed and missing elements were interpolated using the Delaunay triangulation. The exemplary fluorescence signals (EEMs) of peppermint extract before and aft er scatter corcection are shown in Fig. 1.

Prior to the construction of a calibration model, it ts important to analyze the data structure to foresee any possible difficulties at the modeling stage. In particular. data homogeneity. induding the detection of outłying samples and possible measurement errors that can negarive-ly influence construction of a least squaies model, is evaluated 125). Data multwariate exploratory methods are required in order to uncover the structure of the EEMs. Principal component analysis. PCA (26). is frequently used to compress and visualize multivariate data. In PCA. the original data variables are described by a set of new variables, which are called principal components. PCs. They maximize the descrip-bon of the data variance. Data structure is visualized on score and load-ing plots that reveal similarities among the samples and variables. respectively.

To predict the TAC values of samples from their EEMs, it is necessary to construct a mułtivariate calibration model. Assuming a linear rela-tionship between the TAC value and fluorescence spectra, the partial least squares regression. PLS. is a straightforward choice |20). Another

a

C

C

200

MO

Em««kyi (nm) ⁴⁶⁰

E*crtal«or (nm)

Emiunr (nm]

Ex dla bon (nm]

Fig. 1. Exemplary flunnimn ugrali (EEMi] for a prpprrminl ritTart: a] raw ugnal and bj łignjl afrrr wanfffomrrrton.

Strona 88