INTERNATIONAL JOURNAL OF FOOD ENGINEERING Discrimintion of Wines Produced from Cbernet Suvignon Grpes Treted with Aqueous Ethnol Post-Bloom Using n Electronic Nose Amnd Mrtin 1, Kumr Mllikrjun, 2 nd Bruce Zoecklein 3 * Enology-Grpe Chemistry Group, Virgini Tech, Blcksburg Virgini, USA, 24060-0418. ABSTRACT Wine discrimintion nd nlysis is typiclly done through chemicl nlysis nd sensory evlution by trined pnel. Both of these methods re proven to be successful in wine discrimintion, but require extensive preprtion, time nd money. The electronic nose is n objective, rpid-nlysis tool tht hs been used in the food industry for number of pplictions. The purpose of this study ws to determine if n electronic nose cn ccurtely discriminte between Cbernet Suvignon (Vitis vinifer L.) wines mde from grpes tht hve received different pre-hrvest but post-bloom spry tretments to enhnce growth. Aqueous ethnol, which hs been shown to impct fruit mturity, ws spryed on the grpe clusters t 13 weeks post bloom in different concentrtions (control, 5% nd 10% v/v). Chemicl nlysis ws ble to ccurtely discriminte between the wines produced from these grpes. Tringle difference testing by consumer pnel ws not ble to differentite between the different tretments. The electronic nose dt ws ble to ccurtely identify the control group nd the 5% EtOH tretment 90% of the time. Plcement of the 10% EtOH group ws only 13% correct. The results show the promising potentil for n electronic nose to discriminte between control nd treted wine smples.
Keywords: electronic nose, wine discrimintion, sensory evlution, Cbernet Suvignon 1 Grdute Student, 2 Associte Professor, nd 3 Professor nd Hed, Enology- Grpe Chemistry Group, Virgini Tech *Corresponding uthor [Fx 540 231 9293; emil bzoeckle@vt.edu] INTERNATIONAL JOURNAL OF FOOD ENGINEERING Vol. 4: Iss. 2, Article 14. 2008.
INTRODUCTION Reserch indictes tht queous ethnol vpor nd spry cn increse the nthocynin concentrtion nd ripening in tomtoes nd crnberries, respectively (Frg nd others 1992; Beuleiu nd Sltveit 1997). It hs lso been reported tht queous ethnol spryed t 8-13 weeks post bloom on Cbernet Suvignon (Vitis vinifer L.) grpe clusters incresed the nthocynin content in berry skins, juice, nd wines (El Keremy nd others 2002). Anthocynin synthesis is relted to rom nd flvor production nd the formtion of glycosides, which suggests tht fctors effecting nthocynin development my lso ffect rom nd flvor development. This is significnt to grpe nd wine production, becuse rom nd flvor re identifying fctors in grpe mturity. Wine qulity is subjective judgment tht depends on the degree to which the wine is stisfying nd blnced nd reflects the chrcter of the grpe. Therefore, grpe mturity is n importnt wine qulity prmeter. Discrimintion nd qulity nlysis of wine is completed to ensure tht vlue product is produced. Current methods for wine discrimintion nd qulity nlysis include chemicl nlysis nd sensory evlution from trined sensory pnel. Chemicl nlyses cn ccurtely discriminte between wines, but cn be very time intensive nd expensive (Burtti nd others 2007). The wine industry currently uses ph, lcohol, titrtble cidity, ºBrix, trtric nd mlic cids, totl glycosides, phenol free glycosides, totl phenols, color intensity nd hue, totl nthocynins, pigment cofctors nd polymers in order to nlyze wine nd determine the qulity. Sensory evlution is nother wy to discriminte between wines nd ssess the qulity of the wines. Unfortuntely, sensory nlysis often requires specilly trined sensory pnel, requires long preprtion time nd specilist in sensory evlution to nlyze the dt (Burtti nd others 2007). Sensory pnels re subjective, mking it difficult to replicte dt, nd it cn be difficult to correlte the dt received from sensory pnel with chemicl nlysis dt. Becuse of these drwbcks with current wine discrimintion techniques, identifying n objective, rpid-nlysis technique for wine discrimintion would sve the wine industry time nd money (Grci nd others 2006). The electronic nose is reltively new technology tht hs gined populrity in the food industry for number of different pplictions (Lozno nd others 2005; Grci nd others 2006; Moens nd others 2006) from food nlysis (Ntle nd others 1997) to bioprocess monitoring (Bchinger nd Mndenius 2001). It hs multisensor rry tht is used to mesure rom compounds much like the humn olfctory system. The unique feture of the electronic nose system is tht its response tkes into ccount ll the chrcteristic fetures (chemicl nd
physicl) of smple, but does not provide informtion bout the composition of the complex mixture (Hi nd Wng 2006). Becuse of this, the electronic nose system hs been proposed s method of wine discrimintion, where voltile compounds re most importnt for discrimintion nd nlysis. Reserch hs been done using n electronic nose with metl-oxide sensors to clssify different types of wine with different denomintions of origin (Burtti nd others 2004). Grci nd others (2006) lso used metl-oxide electronic nose to clssify four different wines of the sme vrietl which come from the sme cellr. In fct, over the pst few yers, numerous ttempts hve been mde to utilize these chemosensory systems for wine nlysis nd discrimintion (Gudrrm nd others 2001; Kllithrk nd others 2001; Penz nd Cssno 2004; Lozno nd others 2005; Grci nd others 2006; Burtti nd others 2007). For this study, wines mde from Cbernet Suvignon grpes tht received three different queous ethnol tretments during mturity were tested vi chemicl nlysis, sensory evlution, nd electronic nose evlution. The overll objective of the study ws to determine if qurtz microblnce electronic nose system ws cpble of discriminting between these three tretments s ccurtely s sensory pnel. Sensory nd chemicl prmeters were lso compred. The hypothesis of the study ws tht the electronic nose will be ble to discriminte between the three different tretments, similr to tht of the chemicl nd sensory nlysis. MATERIALS AND METHODS Experimentl Design Six vines per tretment were rndomly selected within Cbernet Suvignon (Vitis vinifer L.) vineyrd in northwestern Virgini t the Winchester Field Reserch Fcility, Winchester, Virgini. Tretments consisted of control (wter) nd queous ethnol (5% or 10% v/v) spryed on the clusters 13 weeks post bloom. Ech tretment hd 16 vines for totl of 48 Cbernet Suvignon vines vilble for the experiment. Three fermenttion replictions were prepred for ech tretment for totl of 9 experimentl units. Wine chemicl nlyses utilized by the wine industry were completed for this study. Electronic nose smpling ws conducted 5 months post fermenttion for wine discrimintion nlysis nd to give bsis of comprison for the wine tht ws presented to the sensory pnel. Winemking Approximtely 87kg of fruit ws hrvested per tretment, with n verge of 1.8kg of grpes per vine. Fruit ws trnsported to the Virgini Tech Reserch Winery, Blcksburg, Virgini nd ws kept t 7 C until redy for processing. Discolored nd soft fruit ws removed from ech lug in order to improve the
qulity of fruit to be processed. The grpes were destemmed/crushed with 70% berry brekge using Wottle type-a2 destemmer/crusher (Figure 1). Pectinse (ColorX, Scott Lbs, Petlum, CA) t rte of 100mL/ton ws dded post-crush. Twenty-two kg of ech tretment ws trnsferred to seprte cylindricl tnks nd cold soked t 7 C for four dys with n ddition of 250mg/L dimethyl dicrbonte (DMDC) (Velcorin, Byer Corportion, Pittsburgh, PA). Must ws mixed once dy during the cold soking. Post cold sok, ech lot ws inoculted with 0.24g/L Scchromyces cerevisie strin ICV-D254 (Scott Lbs, Petlum, CA). Fermenttion ws conducted in cylindricl tnks t 27 C with hnd cp punching three times dily, five hours prt (Figure 2). At dryness ( 2.0g/L residul sugr), wines were dejuiced with n ddition of 250mg/L dimethyl dicrbonte, cold settled t 7 C for 24 hours, nd rcked into 3gl glss crboys in n nerobic environment. An ddition of 40mg/L sulfur dioxide ws mde post-rcking. The wines were rcked into 1gl continers prior to sensory evlution nd electronic nose evlution, nd they were stored t constnt temperture of 7 C. Figure 1. Wottle type-a2 destemmer crusher used for Cbernet Suvignon berry destemming nd crushing Figure 2. Fermenttion replictions for Cbernet Suvignon wine production Wine Chemicl Anlysis The commercil hrvesting stndrds for Cbernet Suvignon were utilized for this study. Brix ws determined using n Americn Opticl model 10419 temperture-compensting refrctometer nd ph with Fisher (Pittsburgh, PA) Accument model 20 ph/conductivity meter. Titrtble cidity ws determined by titrtion with NOH to n end-point of ph 8.2. Totl glycoside concentrtion ws determined s described by (Ilnd nd others 1996), nd modified by (Zoecklein nd others 2000) with 200mg polymeric reverse-phse
extrction crtridges using Strt X HLB (Phenomenex, Torrnce, CA). Phenol-free glycosides were nlyzed s described by Zoecklein nd others (2000) using Osis HLB hydrophilic, lipophilic blnce (Wters, Milford, MA). Totl phenols (A 280nm 4), color intensity (A 520nm + A 420nm ), color hue (A 520nm /A 420nm ), totl nthocynins (20 x A 520nm ), nd polymers were estimted spectrophotometriclly (Genesys 5, Spectronic Instruments Inc., Rochester, NY). Electronic Nose Anlysis Five replictions of ech wine were evluted by the electronic nose system s described by the QMB6 user s mnul (Perkin-Elmer 1999). The HKR Sensorsystems QMB6 system equipped with six qurtz crystl bsed sensors (Figure 3) used ws connected to n utomtic hedspce smpler (Model HS-40, Perkin-Elmer LLC, Norwlk, CT). The system, whose sensors coted with vrying degrees of ffinity to polr compounds (polr to non-polr) nd oscillted t 10MHz, used 21mL hedspce vils nd nitrogen s the crrier gs. During the electronic nose evlution, smple hedspce ws generted nd pssed through sensor chmber by pressurizing the hedspce with the nitrogen crrier gs. In order to obtin mximum sensor response, operting prmeters for the electronic nose were optimized for the wine smples. Equilibrium time ws set t 20 min, sensor temperture t 40 C, nd smple temperture ws set t 56 C. Figure 3. Electronic nose system used for Cbernet Suvignon nlysis
Sensory Evlution Wines were evluted 5-months post fermenttion t the Food Science nd Technology Sensory Lbortory, Virgini Tech, Blcksburg, Virgini using tringle difference testing. Two tests were conducted per pnelist per session concerning rom nd flvor. For ech session, pnelists were given 10 minutes to determine n rom difference nd 10 minutes to determine flvor difference. Pnelists were given orl instructions t the beginning nd written instructions during ech testing. In the tringle test, ech pnelist ws given three smples of wine with rndom code numbers to nlyze. Two of these smples were the sme nd one ws different. The pnelists were sked to identify the different wine bsed on rom. Once those smples were tken wy, ech pnelist ws given three new smples with different rndom code numbers for the flvor testing. The pnelists were once gin sked to identify the wine tht ws different bsed on flvor. Smples were presented to the pnelists under red light to ensure tht color vritions would not effect their decision. Sttisticl Anlysis Multivrite discriminte nlysis ws performed on the dt obtined from this study using the sttisticl softwre pckge PC-SAS (SAS Inc., Cry, NC). The GLM procedure ws run to generte ANOVA tbles, generte plots nd run t-tests on the dt. In ddition STEPDISC, CANDISC nd PRIN COMP were used to identify significnt vribles, cnonicl discriminte nlysis nd principle component nlysis, respectively. The PROC STEPDISC in SAS ws lso used to rnk sensor dt bsed on the contribution to the discrimintory power of the system. In ddition to SAS, the electronic nose dt ws evluted using QMBSoft v. 1.22 (QMB6 softwre). Evlution of the sensory dt ws completed using sttisticl tbles locted in Meilgrd nd others (1999). RESULTS AND DISCUSSION Wine smples were ssessed using chemicl nlyses, sensory evlution, nd electronic nose evlution. The dt obtined from this work ws used to mke comprison between the electronic nose nd the other methods of wine nlysis in order to determine if the electronic nose cn be used s discrimintory tool for wine nlysis. The voltiles tested were from Cbernet Suvignon wines tht received queous ethnol tretments (0%, 5% nd 10% v/v) during grpe mturtion.
Wine Chemicl Anlysis Ech tretment of wine ws evluted using the sme set of chemicl nlyses including ph, percent lcohol, titrtble cidity, totl nd phenol-free glycoside concentrtion, color hue, color intensity nd totl nthocynins. Tble 1 provides summry of the results tht were obtined from the chemicl nlyses. Principle component nlysis ws completed in order to verify cluster structure within the dt. The plot in Figure 4 shows distinct seprtion in the different tretment groups. Tble 1. Summry of wine chemicl nlysis results for ech tretment Trt Lot ph TA Alcohol PFGG TGG Color Hue TAC (g/l) (%) (μm) (μm) Intensity 0 1 3.86 6.03 12.7 118.45 1616.88 4.36 0.62 20.87 0 2 3.88 6.18 12.7 122.28 1738.75 4.31 0.61 23.68 0 3 3.94 6.05 12.7 n 1 1438.75 4.17 0.65 25.45 5 1 3.97 6.48 12.5 105.03 1588.75 4.52 0.63 26.05 5 2 3.97 6.16 12.4 77.24 1707.50 4.83 0.68 25.58 5 3 3.98 6.80 12.4 122.28 1738.75 4.22 0.66 23.01 10 1 3.96 6.46 12.3 104.08 1504.38 3.83 0.64 22.64 10 2 3.93 6.33 12.3 105.03 1532.50 4.09 0.66 22.59 10 3 3.94 7.16 12.2 84.91 1298.13 3.64 0.63 22.90 Trt Tretment TA Titrtble Acidity PFGG Phenol-free Glycosides TGG Totl Glycosides TAC Totl Anthocynins 1 PFGG redings for control tretment lot three were not vilble Figure 5 gives visul representtion of the dt results shown on reltive scle. No sttisticl significnce ws found between the three tretments bsed on titrtble cidity, totl nd phenol-free glycosides, hue or totl nthocynins. Results from ph displyed sttisticl significnce between 0% EtOH nd 5% EtOH. Percent lcohol ws sttisticlly significnt for ll three tretments, nd color intensity resulted in sttisticl difference between 5% nd 10% EtOH respectively. Becuse there were three vribles tht resulted in significnt difference between the tretments, the CANDISC procedure ws run using SAS (SAS Inc., Cry, NC) to determine which two vribles were the most significnt in seprting the tretments. Percent lcohol nd ph were found to be the most significnt vribles in determining the seprtion of the tretments.
1.5 1 0.5 Prin2 0-2.5-2 -1.5-1 -0.5 0 0.5 1 1.5 2 2.5-0.5-1 -1.5 Prin1 0 5 10 Figure 4. 11 Principle component nlysis plot of chemicl nlysis dt for Cbernet Suvignon wine smples mde from grpes treted with 0%, 5% nd 10% queous ethnol solution Reltive Comprison of Chemicl Anlyses Reltive Scle b b b c b b ph TA Alcohol PFGG TGG Intensity Hue Anthocynins Chemicl Anlyses 0 5 10 Figure 5. Comprison of chemicl nlyses on reltive scle bsed on tretment
2 1.5 1 0.5 0-8 -6-4 -2 0 2 4 6 8-0.5-1 -1.5-2 0 5 10 Figure 6. Cnonicl plot of chemicl nlysis dt for Cbernet Suvignon wine smples mde from grpes treted with 0, 5 nd 10% queous ethnol solution A cnonicl plot, seen in Figure 6, ws creted bsed on ph nd percent lcohol tht effectively shows the seprtion of the three tretments. Cross vlidtion of this informtion shows tht the chemicl nlyses were 100% effective in discriminting between the three different tretments. Sensory Evlution The first sensory test ws completed for the comprison of the control tretment (0% EtOH) nd the 5% EtOH tretment. The following sttisticl prmeters were set for the evlution: p d = 0.30, α = 0.05, nd β = 0.20. A totl of 48 responses were collected for this comprison, using tringle test for rom nd tste. A totl of 22 correct responses were needed in order for the two tretments to be sttisticlly significnt. A summry of the responses from this sensory evlution cn be found in Tble 2. Fourteen correct responses were received for rom, nd 15 correct responses were received for tste. Therefore, the 0% EtOH tretment nd the 5% EtOH tretment were not found to be significntly different.
Tble 2. Sensory evlution of control nd 5% EtOH Correct Incorrect 0% - 5% AROMA 14 34 0% - 5% TASTE 15 33 Tble 3. Sensory evlution of control nd 10% EtOH Correct Incorrect 0% - 10% AROMA 14 19 0% - 10% TASTE 11 22 The second sensory evlution ws completed for the comprison of the control group (0% EtOH tretment) to the 10% EtOH tretment. The following prmeters were set for the evlution: p d = 0.30, α = 0.05, nd β = 0.30. A totl of 33 responses were collected for this comprison, using tringle test for rom nd tste. A totl of 17 correct responses were needed in order for the two tretments to be significntly different. A summry of the responses from this sensory evlution cn be found in Tble 3. Fourteen correct responses were received for rom, nd 11 correct responses were received for tste. Therefore, the 0% EtOH tretment nd the 10% EtOH tretment were not found to be significntly different. Electronic Nose Evlution Ten replictions of ech wine smple tretment were evluted using the QMB6 sensorsystem. Dt ws first nlyzed using the QMB6 softwre (QMBSoft v. 1.22). A visul representtion of this nlysis cn be seen in Figure 7. From this plot, cler seprtion between the control group nd the treted smples cn be seen. The electronic nose ws successful in discriminting between the control smples nd the smples tht received tretments. The CANDISC procedure ws performed on the results from the electronic nose redings in order to produce cnonicl plot. From this plot (Figure 8), seprtion cn be seen between the control group nd the treted groups (5% nd 10% EtOH).
Figure 7. Projection plot using QMBSoft of control, 5% nd 10% tretments from the discriminte nlysis of sensors of the HKR Sensorsystems QMB6 system The cross-vlidtion for this informtion (Tble 4) identifies the number of observtions nd the percent clssified into ech tretment by the electronic nose. This tble shows tht the electronic nose ws successful in ctegorizing the control group nd the 5% EtOH tretment group 90% of the time. However, the ctegoriztion of the 10% EtOH tretment ws only successful 13% of the time. It is possible tht the electronic nose ws recognizing the 5% nd 10% EtOH tretments s the sme smples nd plcing them in the sme ctegory. This cn be clerly seen in Figure 7, where the 5% EtOH tretment ppers to be subset of the 10% EtOH tretment. Bsed on this interprettion, the electronic nose ws ble to correctly plce the 5% EtOH tretment 100% of the time, nd the 10% EtOH tretment 80% of the time.
3 2 1 0-6 -4-2 0 2 4 6-1 -2-3 Control 5% 10% Figure 8. Cnonicl plot of electronic nose dt for Cbernet Suvignon wine smples mde from grpes treted with 0, 5 nd 10% queous ethnol solutions Tble 4. Cross vlidtion of the discriminte nlysis of electronic nose dt for wine smples mde from Cbernet Suvignon grpes treted with 0%, 5% nd 10% v/v queous ethnol solution Number of Observtions (Percent) Clssified into Trt From Trt 0 5 10 Totl 0 9 (90) 1 (10) 0 10 (100) 5 0 9 (90) 1 (10) 10 (100) 10 2 (20) 5 (50) 3 (30) 10 (100) Totl 11 (37) 15 (50) 4 (13) 30 (100) Priors 0.33 (3) 0.33 (3) 0.33 (3) Error Count Estimtes for trt 0 5 10 Totl Rte 0.1 0.1 0.1 0.3 Priors 0.33 0.33 0.33
The electronic nose is unique in tht it gives n overll objective comprison between smples. It is not designed to discriminte between smples bsed on the intensity or concentrtion of prticulr element, but rther the responses of ech of the sensors form chrcteristic recognition pttern for tht prticulr smple. This my explin why there seems to be no direct correltion between the electronic nose dt nd the chemicl dt. The electronic nose will be more sensitive to the overll rom of the wines, nd would therefore correlte well with phenol-free glycosides (PFGG) in the wine. The PFGG results did not show significnt difference. A significnt difference between the smples ws found bsed on the percent of lcohol in the wine; however, the electronic nose did not discriminte between different percentges of lcohol s result of normliztion. The electronic nose my hve been sensitive to the difference in ph tht my hve impcted voltility. From the results of the electronic nose there ws distinct difference between the control group nd the treted smples. Bsed on this informtion, the ppliction of queous ethnol post-bloom resulted in chnge in the wine produced. CONCLUSIONS The purpose of this study ws to determine if electronic nose technology could ccurtely discriminte between wines mde from Cbernet Suvignon fruit tht received different queous ethnol tretments post-bloom. The electronic nose dt ws compred to chemicl nlysis dt nd results from sensory nlyses. The chemicl nlyses were ble to effectively discriminte between the three smples (control, 5% nd 10% ethnol spry tretments). These results re importnt becuse they verify tht the wine smples being evluted cn be discriminted nd grouped. However, discrimintion ws due to only three significnt prmeters: color intensity, percent lcohol nd ph. Despite the fct tht there ws cler seprtion between groups s determined by the chemicl nlysis, the sensory pnel ws not ble to discriminte between the control group nd the treted wine smples. The electronic nose ws successful in grouping the control group seprtely from the ethnol spry tretments evluted in this study. The treted smples were grouped in the sme re. Bsed on these results, it ws found tht the electronic nose cn be used s discrimintory tool for control nd treted wine smples under the conditions of this study. Further chemicl nlysis should lso be completed in order to better identify significnt prmeters tht differ mong the smples, including specific voltile components.
REFERENCES Bchinger T, Mndenius C-F. 2001. Physiologiclly motivted monitoring of fermenttion processes by mens of n electronic nose. Engineering Life Science 1(1):33-42. Beuleiu JC, Sltveit ME. 1997. Inhibition or promotion of tomto fruit ripening by cetldehyde nd ethnol is concentrtion dependent nd vries with intil fruit mturity. Journl of Americn Society for Horticulturl Science 122:392-8. Burtti S, Bllbio D, Benedetti S, Cosio MS. 2007. Prediction of Itlin red wine sensoril descriptors from electronic nose, electronic tongue nd spectrophotometric mesurements by mens of Genetic Algorithm regression models. Food Chemistry 100(1):211-8. Burtti S, Benedetti S, Scmpicchio M, Pngerod EC. 2004. Chrcteriztion nd clssifiction of Itlin Brbr wines by using n electronic nose nd n mperometric electronic tongue. Anlytic Chimic Act 525:133-9. El Keremy A, Chervin C, Souquet J-M, Moutounet M, Monje M-C, Nepveu F, Mondies H, Ford CM, vn Heeswijck R, Roustn J-P. 2002. Ethnol triggers grpe gene expression leding to nthocynin ccumultion during berry ripening. Plnt Science 163(3):449-54. Frg KM, Plt JP, Stng EJ. 1992. Ethnol enhnces the effectiveness of ethephon on nthocynin production in crnberry fruits in the field. Journl of Americn Society for Horticulturl Science 122:392-8. Grci M, Aleixndre M, Gutierrez J, Horrillo MC. 2006. Electronic nose for wine discrimintion. Sensors nd Actutors B: Chemicl 113(2):911-6. Gudrrm A, Fernndez JA, Iniguez M, Souto J, de Sj JA. 2001. Discrimintion of wine rom using n rry of conducting polymer sensors in conjunction with solid-phse micro-extrction (SPME) technique. Sensors nd Actutors B: Chemicl 77:401-8. Hi Z, Wng J. 2006. Electronic nose nd dt nlysis for detection of mize oil dultertion in sesme oil. Sensors nd Actutors B: Chemicl 119(2):449-55. Ilnd PG, Cynkr W, Frncis IL, Willims PJ, Coombe BG. 1996. Optimiztion for the methods for the determintion of totl nd red-free glycosyl glucose in blck grpe berries of Vitis vinifer. Aust J Grpe Wine Res 2:171-8. Kllithrk S, Arvnitoynnis IS, Kefls P, El-Zjouli A, Soufleros E, Psrr E. 2001. Instrumentl nd sensory nlysis of Greek wines: implementtion of principle component nlysis (PCA) for clssifiction ccording to geogrphicl origin. Food Chemistry 73:501-14.
Lozno J, Sntos JP, Horrillo MC. 2005. Clssifiction of white wine roms with n electronic nose. Tlnt 67(3):610-6. Moens M, Smet A, Nudts B, Verhoeven J, Ieven M, Jorens P, Geise HJ, Blockhuys F. 2006. Fst identifiction of ten cliniclly importnt microorgnisms using n electronic nose. Letters in Applied Microbiology 42:121-6. Ntle CD, Mcgnno A, Dvide F, D'Amico A, Polesse R, Boschi T, Fccio M, Ferri G. 1997. An electronic nose of food nlysis. Sensors nd Actutors B: Chemicl 44:521-6. Penz M, Cssno G. 2004. Chemometric chrcteriztion of Itlin wines by thin-film multisensors rry nd rtificil neurl networks. Food Chemistry 86:283-96. Perkin-Elmer C. 1999. Opertors mnul for the chemosensory system QMB6 / HS40XL. Norwlk, CT: HKR Sensorsysteme GmbH. Zoecklein BW, Dougls LS, Jsinski YW. 2000. Evlution of the Phenol-Free Glycosyl-Glucose Determintion. Americn Journl of Enology nd Viticulture 51:420-3.