Scribd Logo
Fazer o upload

Bem-vindo(a) ao Scribd!

  • Aw Gabriel

    Enviado por

    Nils Huaman Castilla
    6 visualizações8 páginas
    Tecnologia de alimentos

    Título original

    AwGabriel[1]

    Direitos autorais

    © © All Rights Reserved

    Formatos disponíveis

    PDF ou leia online no Scribd

    Você considera este documento útil?

    Este conteúdo é inapropriado?

    Denunciar este documento
    Tecnologia de alimentos

    Direitos autorais:

    © All Rights Reserved
    Baixe no formato PDF ou leia online no Scribd
    Sinalizar o conteúdo como inadequado
    6 visualizações8 páginas

    Aw Gabriel

    Enviado por

    Nils Huaman Castilla
    Tecnologia de alimentos

    Direitos autorais:

    © All Rights Reserved
    Baixe no formato PDF ou leia online no Scribd
    Sinalizar o conteúdo como inadequado
    de 8
    Available online at vavw.sciencedirect.com -" ScienceDirect Food Chemisiry 18 (20081 1106-1113 Food Chemistry Analytical Methods Estimation of. water activity from pH and °Brix values of some food products _ Alonzo A. Gabriel * Deportment of Food Science anud Nusriion, College of Home Econoonics, University of the Phtippnes, Dian. Quezon City Pllipines Received 8 June 2007; received fn revised form 31 October 2007: weeepted 26 November 2007 Abstract In this study, a predictive modal forthe estimation of water activity (uf) as a function of pH {2,00-8.00) and *Brix (0-82.00) values of simulated food solutions (SFS) was developed, through response surface methodology. Response fit analyses resulted in a highly sg. nificant (pH < 0.0001) square root polynomial modet that:can predict «2°° of SFS in terms of pH and *Brix values within the defined variable ranges, The linear, quadratic and interactive inflienees of pH aid *Brix on af" were all sighticant (pH < 0.0001), Medel val idations in SFS and in a number of actual food systems showed that the model had acceptable predictive performance, us indicated by the calculated accuracy and bias indices, © 2007 Elsevier Ltd. All rights reserved Keywors: “Brix: pl: Predictive modelling; Response surface methodology (RSM); Soluble solids: Water activity 1. Lntroduetion Water activity (ay) is a measure of the amount of water available for chemical reactions, as well as microbial growth in food (Belitz & Grosch, 1999; Jay, 2000; Rock- land & Beuchat, 1987; Troller é&:Christian, 1978). Further- more, the mode and severity of food processing for several commodities may be highly dependent on the ay of the product (Zapsalis, 1985). Therefore, measurement of «ty is essential to the food industry, since it plays a vital vole in addressing the needs for product stability. quality mainte- nance and sustaining the safety of food throughout its shelf-life, Sucrose is a common ingredient of many food products and used as sweetener or a humectant (Triebold & Aurand, 1963; Troller & Christian, 1978). Shelf-life sta bility of jams, marmalades, fruits in syrups, and other sweetened food products rely on the abilily of sucrose to Preset address: Laborutory of Food Microbiology and Hlygiene, Giudnute School of Biosphere Science, Hireshima University, 44 Keapamiyama, Higashi Mieoshim 739-8538, Japa, Eanail addesses: aloncogabrickhiroshinna-u.:jp, alonro,gsbvikia coped © 2007 Ehevier Led, All rights reserved. ov 308-8146 - see Front mit {doi 10. 1016) Foodchem.2007, reduce the dy {0a level where microbial growth and unwanted chemical reaction rates are slowed down ‘One major factor that influences dis the concentration and-type of solute present in the food system (Fennec 1996; Nielsen, 1994). Generally, by simply following Rao- ul’ Law of mole fraction, increase in the amount of solute in esystem shall ideally result in a predictable decrease in 4 (Fennema, 1996; Holtzclaw & Robinsons. 1988; Trotler & Christian, 1978), Jay (2000), however, cited that many solujes, includitig the disaccharide sucrose, do not follow Reoult’s Law. Interactions of several food properties may possibly explain such a phenomenon. For example, addition of acids in a system containing sucrose causes sugat inversion (Andiews, Godshall, & Moore 2002; Ben- nion, 1985; MeWilliams, 1993), Invert sucrose has been reported to have a greater ay lowering effect on foods than sucrose alone (Fennema, 1996). Despite being one of (he more important food:param- eters that aflect food quallity and safety, micro- to med- iium-seale food processors are not able to allord ay meters. Prices of the east expensive a meter models an be as much as US$2000 (Cole-Parmer Instrument Company, Water activity meter systems,, 2007; Decagon sans an AA, Gabvict Food Chemisty 108 (2008) 110854113 Devices. Water activity for food science: assuring safety & governmental compliance. 2007; Novasina, Water activity: For applications in the food & cosmetic indus tuy.. 2007), Furthermore, operation of ay meters may also be a concern, since stakeholders often lack man- power with sulficient technical knowledge. This study, therefore, tried addressing this gap. by developing a mathematical model capable of predicting the a at fixed lemperature (233°) of some foods, from easily measured, pertinent physicochemical food properties, namely pH. and °Brix. values. The study established a predictive model in simulated food solutions (SFS), which com tained varying levels of water, sucrose and avid. The pre dictive performance of the established model was assessed through validations using a different set of SFS and various actual food systems 2. Materials and methods 2.1, Simulated food solutions The SFS were formulated based on the data supplied to and processed using the Design Expert Version 7.0.3 soft: ware package (Statease, Minneapolis, MN). A Central Composite Rotatable Design (CCRD) was applied, to determine the appropriate combinations of various levels of pH and °Brix. Table | presents the coded and uncoded SFS formulations that resulted from the CCRD. The assigned °Brix values per SFS were achieved by dissolving food grade p-sucrose (Ajax Finechem, Australia) in de-ion- ised distilled water, The’ desired pH value per SFS was adjusted using 5N HCI (Himedia, Munsbai, India) o: 8N NaOH (Himedia). Freshly prepared solutions were immediately subjected to ay analyses. Table 1 Rottable central composite design use! in the Formnilatien of simulated Food sofoions Expermental one jocks Coded variable” Uncoded variable combinations pH 1.00 values are indicative of model-overestima- see meaminl MIS"C wig a Nowa" alge Giomine ions, A By-value of 1.00 signifies that the ®a2°°° and the smitvrtand) water activiy meter: Valuen ave prsented we uetoee a *ag’'© are in perfect agreement, Since the By value does the tials stardnrd deviation nol provide an indication of model predictive accuracy, the A; wis also calculated bates sere an ad ' rests, iean cleurly be seen thai both predictive variates {tal} influenced the measured response, SES with higher *Brix (3) waives had tower ae than those with lower *Brix ata fixed pli level. The eflect of diflering pH on the a, of SFS with equal °Brix was also demonstrated. The ificance of these influences and the possible existence of non-tinear and interactive predictive’ variable influences were further explored and are discussed in the following sections. The influences of pH and *Brix on SFS a2" were quanti- tatively and qualitatively characterised by fitting the * obtdined results (Table 2) into’a second-order polynomial model (Eq. {1)}. This model was used to account for possible non-linear relationships between the predictive and response variables (Mendenhall & Sincich, 1996). Hu (1999) explained that lower degree polynomial models, such as those with interaction and quadratic terms, ate appropriate to ade- quately describe food processes. Results of the Box-Cox Power transform analysis (Design Expert 7.0.3, Statease) of model fit on Eq. (1) (current = 1,00, best 4 = 2.48) how- “sever suggested thal the results obtained from the study had betler fit on Eq. (4) (current &= 1,00, best & = 1.22). Ay 10 Take note that the only difference between Egs.(2).and (3) is that dy value measures. the absolute difference between thea? € and the*az""©, The A/takes values > 1.00, where greater 4;values are indicative of less model predictive accuracy, An Ar= 1.00 is also an indication of perfect agree- ‘ment between the predicted and actual values. Thus for cases where PaiS"© > *a%*"C, the calculated. 4, and By values will be equal, since both fogie and llogio| of (Fa *a2"©) result in the same positive number. However, when "ay < *aZ, the logy and flogio| of the quotient of ‘the predicted and actual “ag*"" values shall result in numbers with the same value but opposite signs; henee Ay. By Graphical cotiiparisoins of the predicted and actual calew lated ai" were algo done by plotting "422" against §a2"¢ ‘The line of equivalence (LOE) was traced, to indicate the region of the plot where fai" = *a85°°. The LOK is thelline with an equation y= x and diagonally bisects the plot into two equal regions. A point falling on the LOE has Ar= By= 1.00, hence perfect agreement between Par ©. ¥? = fy + (ri) + falta) + Missle: Xan) + Biv alad) and “ais © Points falling-above (4y= B;> 1.00) and below °F Bacal) ' (4,> 1.00, B< 1.00) have been overestimated and underes- Vy feats : timated by the model, respectively. The farther the point ‘The results of the second order model fitting on Eq, (4) are from the LOE. the greater the A,and Byvalues will be. Thus * presented in Table 3. The Fisher F-test results with very tow the positions of the plotted points, with respect to the LOE, _p-valués demonstrated the vety high significance of the may also be used as bases in the evaluation of the perfor- model and conveys that the predictive variables, pIT aid (4) ‘mane of the predictive model. “Brix, can be used to reliably predict the response variable, o38°¢ (Adinarayana & Ellaiah, 2002; University of California 3. Results and discussion al Los Angeles, 2006). In Table 3, itis also shown that the . predictive variables have non-linear and interactive influ- SJ. Predictive model fitting and analysis ‘enees on the response. Thus, utilisation of model Eq. (a) ‘was deemed appropriate. Further, the high lack-of-fit F- Table 2 presents the a," measured from the different value implied that the obtained results fitted well-to the is SFS with varying pH and *Brix combinations, With these model equation. AA. Gabriel! Food Chenistry 108 (2008) 11061113 Table 3 Analysis of variance for the response surface model Sowee Sum oFsquaies _Df_ Mean square Model 210x103 54010 pil @3x19% 1 639x10" <0.0001 s Six Roget! 20x 10 0.0001 s plea S010 1 S30x10 8 rave . b a éeée ¥ 2 oss 0.35] 1 ¢ d . sa : 086 Be 080 50 100 ° e a, 06 art oe ae ) cr ) was Toe Fig. 2. Graphical comparisons between the mest juices and juice drinks c) fruitaectars unl co predits, The nanbers placed nea he points indicate the lence (LOE} denotes the region whore "2! © = = whese A,> 1.2, seated a rates (d) jams, is and in <2 while (he shale cas bot nb of pons thal oineided inthe sameenordinates, The tine bitecting eats pl, the li by the dow bites dite the + 20 swcuat measured "ai" in: 1a) the valkdating simulated Food systems. fb) (iit prodvets, (el: rie rvs, and 1) milk, dasry products ane other foust of prediction error repions ua of a@ © with increasing pH might have been due to the increase in the ions contributed by the concentrated NaOH used to adjust the SFS pH. 3.3. Meulel validation ~~ - A predictive model may only be safely used in decision making when validated (Jagannath & Tsuchido, 200%) Validation is an essential step that reveals the applicable range of a model and the limits of its performance. There- fore this study also dealt with the validation of the devel- coped model, using a set of data obtained from additional lest runs, exclusive from those performed in the elabora tion of the model, as recommended by Ross (1996) and Carrasco et a, (2006). In this new study, the reliability of the developed model was also assessed through validations in real food systems, as,similarly done by Mckiray et a (2000), and Jagannath aiid Tsuchido (2003) Tables 5 and 6 present the calculated values of the per- formance indices for the validation of the additional SES aund real food! systems, respectively. In the validating sot of SES. the calculated Ay values varied from 1.01 to 1.05 (range: 0.04), Moreover, the 4, values caleutated from model validations using actual Food systems ranged from 1,00 to 1.23 (range: 0.23). Ross. Dulaard, rt (2000) reported that predictive models shoutd ideally tiave A= 1.00, which indicates a perfect model lit where the pre- (ed und actual response values are equal. However, Ross et al, (2000) and Carrasco et al. (2006) explained that, typ= ically, the Ay of 4 model increases by 0.10-0.15 units for “every predictive variable in the model. Therefore. a model that forecasts a response from two predictive variables may be expected to have A, values that range from 1.20 to 1.30 (Ross ct al., 2000) oF an equivalent error range of 20- 30%, Therefore based on the resulls obtained from the val idating SFS and actual food systems, the predictive perfor mance of the established model may be considered acceptable. The calculated By values ranged from 0.95 to 1.03 (range: 0.08) and 0.91 to 1.23 (range: 0.32) for SFS and aetual food. systems, respectively. These vilues indicate Uhat the predictive model under- or over-estimated the actual ay of the validating SFS and food systems. Jagan- rath and Tsuchido (2003) reported that higher By values may be expected, if predictive model is validated in con- ditions different from those used in the model establish- ment, Thus, in their study. By values of 41 and 1.60 were considered acceptable after validating a phosphate buffer-based model in actual milk systems. As with the 4, values, slightly higher Bf values were calculated from vali dations in real food systems, compared to those caiculated from the validating SFS. In all validating systems but two samples (sweetened purple yam and condensed milk), where the predictive model overestimated the a, the Ay and By, values were calculated to be equal The discrepancies between the °a:° © and “a9 © enleu- lated from validations in SFS and different actual food A.A. Gabriel Food Chemisry 18 12008) 1106-1112 systems are also illustrated in Fig, 2a-fs In the figure, it can be seen that only three (Fig. 2e) model predictions had% error values greater than 20% (Ay> 1.20). ‘These poinis correspond to the a estimated for raisins, dried cantaloupe and dried pineapples. Nevertheless these points were still within a 30% error (Ap 1.30), and hence can still be considered to have acceptable accursey (Rosset al., 2000), These results agree with the caleu- lated model performance indices. Parallel to the ealet- luted performance indices, graphical validations showed Giat the modet had beter predictive performance in the SFS than in actual. food systems (not emphasised in the figure), Such results may be expected (Jagannath & ‘Tsuchida, 203), since’ the model was established using SFS. The greater differences between Pal" and ai © in aiclual food systems may be duc to the influences af other food components on the a that were not present in the SFSs used in establishing the model. Food comporcnits such a5 proteins, complex and simple carbohydrates, salts anc! other dissolved components might have caused such observations. Despite these variations, results of mathe ‘nuatical validations showed that the established prod model reliably predicted the wy of the validating systems Hence, the predictive model may safely be used sn ealeulat ing the «, of appropriate food systems. 4, Summary and recommendations This study dealt with the development of a predictive response surface modet for the estimation of 3° from PHL and °Brix values of appropriate food systems. The lily significant predictive mode! was devetoped from, simwultied food systems (SFS) but was validated using a separate set of SPS and actual food systems. Results of the validation showed that the developed mode! bad acceptable predictive performance, as assessed by muithe matical and graphical model performance indices. study recommends further validation in other appropriate food products, so as 10 further explore. the predictive capacity and linvitations of the model. Other predictive variables such as temperature and types of solutes may also bbe considered in building up a more accurate and practical predictive model. Acknowledgements This study was made possible by the research grant (Project Number 060603 PNSE) provided by the Office of the Vice Chancellor for Research and Development of the University of the Philippines, Diliman, Quezon City, Philippines. The technical assistance rendered by Joyce B. Pineda and the Pilot Food Plant of the College of Home Econotnies, UP Diliman in conducting the experiments are acknowledged. The assistance of Kirth M. Asis and Joyce B. Pineda in the initial preparation of the manuscript is also recognised AA. Gust! Fond Chenery 18 (2008) 1106. 13 References Adinarayana, K., & Blaigh, 0. (2002). Respnase surface optimization of the critical ined components for the prextiction of slkeline proteose by newly isolated Aacilae sp. Jawinal wf Phurmaes wud Pharvaceation! Scienes, 4), 22-278. Andrews, L.S., Goushall, M.A... Mout, (20021, Sucrose degradation lander model processing eoneitians. Svar of Food Science. 623%. 1621-1624, Aurand, Le Wa & Woods, A. 6. (197%) Favs! Chemistry, Westpork, CT Avi Publishing Company. Ine Fiaianu, 1. C. (1992). Phat efemity of fond processes. Randanental aypects (Vol. 1), Leinhold: Van Nosieand, Baranyi. J, Pin, C., A Ross. T. (1999), Droietive medels. Dnermetinnl Journal of 159-166, Bolte, HD, & Grose, W. (1899). Rad ebeniviry Verlag. Benninn, M. (1983). dunenductory fond: (8th eB). New York: Mactan Box. G. E. P., Hunter. W. G., & Hunter, J. S. (1978). Statistiey for experimenters: Aut hntrchecton 40 desig dat aualysia and model Iidng. New York: Johw Wiley and Sons Carmaca, E., Garcia-Gimeno, R.. Sesetovsky. Ru: Valero. A. Perey Rarer, G. eal. (2006), Predictive woill of Lasteres mone nes growl tate under elfen temspesatures al ais, Fd Serene and Teh Iecrnaviont, 220), 47 Sh ayn, ME, Visnole, Gh. & Garro, 0. (28, Maueling ‘actos growth in sietuinpackuged eopked meat eInuionestored Ht thre temperatures. Pant Micvobielaey. 20, 561-566, Cole-Parmer bisttument Company. Wailer setvity weer systems, 12007) tty ffvorcolaparmer eon’ Aecessi 68.06.17 Decagon Devices. Water ueivity for food stews: ssstring safely. quality Validating and comparing Feo! Microbiol, 4, Beatin: Spinner ind governmental compliance, |2005). = heprf/sw tosugoeomy Accessed 08.607, Vonmenns, 0. 11996). Food cheney Cae aa. NY & Basel: Marcel Dekker tne, Gibson, BUD). The elect of hs ‘existance of vegetative mkroorgunisms, Journal af Applied Bate gy. 36, 05-37, Han, ¥.. Flotos, 31D. Linton, R. HL. Nickson, $. S., & Nokon, PF, (2002). Resporise srrface modeling for the inactivation of Exchetrhis ent OISTANT om green peppers (Cte aunin) By ame gas tavatment. Journal of Food Seine, G713}, UBS. 1193, Holizclay, 11.4 Robinsons, W. (18), Gone Cheat. D.C. Heath & Company. Hu, R. (1999), Fao prude clesigns 4 ronuter-uide attic appre USA: Tectnomie Pulvishing Co, fe supe concentrations en the heat ins Jagunitth, A, & Tsuchide, T. (2008), Waligation of a polynomiat gression mouel the thera inscsivation of Beles sta spe in mil. Lents Applic Microbiology, 37, 399-408 1a}. J M4 (2000), ben food microbiota (6. Singapore: APAC. Poblshes, Moturagts, M.. Drosinos, E. EL, Vaidans, A. Metaopaons. | 20% Development uf x predictive model for the spoilage Hf enoked canal ‘meat products and its vation unger constant and dynamic Lemperatore storage eondivions, Jonna! of Fond Science, 71.157 167 Mtitoy. D. M. faykus. Ld Foegeding, P.M. 12000). Validation nd ‘analysis of modeled predictions of geosmth of Maile cerns ses in boiled tice, Journ! of Fao Protection, 642). 268-272 MeWitliams, M_ (1992). Faud fiowkmentls. New York: Mail Publishing Compan ‘Mendenhall, W.& Sincich, T1996} sera couse tata: Regesvion cai (Sth ec New Jrsey: Peete Ea tert Ine Nielsen, $. (1994). dori ta the chenncal analysis of foals. Bostexs Jones and Barlet Publications Novawina. Woter setvity: For applications in the fou, pharmacies! sand cosmetic industy, (2007), “lneps/www-novasine chi” Accessed 68.0607, RRocktind, L.. & Meuchat. L. C1987). Pater asi: Theory cane! ypiew tens 10 fond. New York: Mareet Dekker Ross, T. (1986), Indices foe peeformance evaltation of predictive mulls Jos of Appel Bacteriolugy, 8. SOV SO Ross, 7, Dulgid, 1, 8 Tienumgoor, 8. (20001. Pradictve meacty: uf nah aul sues asta in Uishery prolels Itermetevaa Jona of Ba Mirrabilo. 62 2M 248, Stallenberger, R. 5, & Binh, G. G. (1998). Sigur cheney, Westport CT: Avi Publishing Casopny. Hue Teint. W., Rodrigo. M..e Martinez, &, (2001), Modeling the combined eilect of pH and temperative on the bet nssstance of Paci Sewratiennaphites spores heated in multicomponent feed estat Smut af Foor! Meectiny, 64 10), VBI 1G3S, Tritbokd. Hl. Se Aarand Le (1963), Fao composition and codes Princeton, NJ: Van Nostrand “Tooller. 1.42 Christian, He (E078). Water activi and food, Ne Acero Press University of California at Los Angeles, Asides Technology Serves Annotated SAS output: Regression analysis. (2006).

    Você também pode gostar