8290614258

7. Opublikowane badania własne

/ One! et of. / falonro 138 (2015) S4-70

rcsponse variable. and y, is the Ith value of the response variable as predicted using the constructed model.

Another parameter that characterizes the explained model yariance is the coefficient of determination. D²:

<3)

where y is the mean value of dependent variabłe.

4. Results and discussion

Firstly, the central composite design was used in order to find the optimal exrraction condirions. As was mentioned eaiiier. the ranges of the temperaturę and sonicarion rime were considered to be in agreement with the limitations of the sonication bath. The second-order model as described by Eq. (1) was fitted to the data and the significance of the regression coefficients was verified using the analysis of yariance, ANOVA. with the type III sum of squares (see Table 2). The assumed level of significance was 0.05 (P-value < 0.05).

From the P-values that are presented in Table 2. one can conclude that the regression coefficient of variable Xj. which is assodated with the time effect, is insignificant (P-value > 0.05), but the regression coefficient that describes its quadratic form is found to be significant (P-value < 0.05). This suggests a nonlinear dependency of the ORAĆ parameter and the extraction sonication time. The finał regression model can be expressed with the fol łowing equation:

y= 119.70+ 10.73x2 + 7.22x?-5.75x₂-5.61x,x₂+f    (4)

The coefficient of determination. D². was equal to 0.835. Because repeatcd measurcments were performed. the residua! yariance can be partitioned further into the yariance related to the pure measure-ment error and the yariance of lack-of-fit with ANOVA. The model that is constructed is adequate and useful because the lack-of-fit was insignificant (P-vaIue > 0.05). The response surface depicted in Fig. 1 shows one saddle point

The largest value of the ORAĆ parameter was found for the highest temperaturę of 80 C and the shorrest time of sonication (3 min). This finding is in agreement with the suggestion of some authors that sonication and a long time of high temperaturę exposure may lead to a degradation of some antkucidant substances [ 30.311 and conse-ąuently to a decrease in the antioxidant capacity values as measured by the ORAĆ assay. The optimal condirions were further yalidared with addirional measurcments. The RMSE yalue was equal to 13.14 TE (7.97%).

The extracrions of antioxidants that were present in the tomato pastes were carried out under optimal condirions. The ORAĆ parameter was evaluated for each extract and the EEM spectra were registered. The fiuorescence and IR spectra for the intact samples were collected at the same time. Raw instrumental signals were then preprocessed prior to multivariate modeling. A visual inspecrion of the signals allowed it to be conduded that the signal-to-noise ratio

Table 2

Rf suita ol fht analysis of vanancr wjfh the type III sum cf squart&.

Smurtt of vjrij!ton	Sum of UjUJRfl	Pniftl of (rffdofu	Mf ah ujiurcf	F-v4Jue	FuKm
	140	1	140	0 070	0 795
I;	1 M m 10^1	1	1 84 . 10'	17<W	< 0 001
	476.37	1	476 37	9 822	0007
*>	102.16	i	102.16	6210	0026
MyK,	2SI 84	i	251 JM	5 191	0039
Model	3 44 . 10’	4	860 00	1831	<0001
Rfitduils	682.16	IS	4S.49

72

O

o

OJ

3

w

&

E

Time (min)

Fis. 1. Rcsponse surface that wai obiamed from the central rnmposite destan. A descriprmn of the rnded parametrrs U presented In Tahlr I.

was at a sarisfactory levd for EEMs and IR spectra and that the noise reduction step was unnecessary. The preprocessing of EEMs con-sisted of removing the Rayleigh and Raman scattering using the routine that was described in detail in |32|. Scattering peaks were detected and removed. Then, the conresponding regions of signals were interpolated using the Delaunay triangulation. Next, each exritadon-emission matrix with the dimensions 26 x 161 was unfokled into a vector with the dimensions 1 x 4186. The finał set of EEMs was presented into a matrix with the dimensions 39 x 4186.

The PCA score plots of the unfolded EEMs for the exrracts of tomato pastes revealed a clustering tendency. Namely. the samples from Italy and Croaria (south of Europę), the projections of which are found along PC 1 axis (see Fig. 2a). can be distinguished.

An analysis of PC 1 loadings was performed after their refolding to an image that corTesponded to the exritation-emission ranges of the EEMs (see Fig. 2b). The intensity of the color of each pixel is proportional to its loading value. Cold colors correspond to Iow values while hot colors indicate high yalues. An analysis of the image allowed for the conclusion that the mixtures that were studied contain Chemical components that exhibit fiuorescence of around 320 nm (excitation around 300 nm). Th esc compounds are found at relatively high concentrations in the samples of tomato pastę from Poland. A similar fiuorescence characterisric is typical for catcchin and epicatechin |33|. Both substances are common antioxidants that are present in tomatoes |34|. A further analysis

Strona 100