W.L.P. Bredie and M.A. Petersen (Editors) Flavour Science: RecentAdvancesand Trends 9 2006 Elsevier B.V. All rights reserved.
521
ChemSensor classification of red wines lnge Dirinck, Isabelle Van Leuven and Patrick Dirinck
Laboratory for Flavour Research, Catholic Technical University St.Lieven, Gebr. Desmetstraat 1, BE-9000 Gent, Belgium
ABSTRACT In this study the hyphenated technique of automated headspace-solid phase microextraction (HS-SPME) and quadrupole mass spectrometry (MS) as a sensing system was used, in combination with on-line pattern recognition algorithms, for classification of red wines. These ChemSensor classifications based on mass fingerprinting were compared with a time-consuming GC-MS analysis, consisting of headspace-solid phase microextraction, identification, semi-quantitative determination of the wine volatiles and principal component analysis (PCA) of the semi-quantitative data. Good correlations could be observed between both techniques. 1. I N T R O D U C T I O N In previous work the ChemSensor system was successfully used for fast aroma characterisation of coffee [1,2]. The aim of the current study was to evaluate the suitability of ChemSensor classifications for fast objective quality evaluation of red wines. For this purpose a good relationship should be obtained between classifications based on: a) fast MS fingerprinting; b) time-consuming GC-MS profiling and c) sensory data from an expert wine panel. A first approach studied a consumer-oriented model system composed of 14 commercial red wines with important sensory differences. In a further stage the suitability of the method was demonstrated by 2 thematic comparisons of Bordeaux wines with, respectively Cabernet Sauvignon and Merlot as dominating grape. However, presentation of these results is beyond the scope of this publication.
2. MATERIALS AND METHODS
2.1. Wine samples A consumer-oriented model system was composed of 14 commercial red wines with important sensory differences and covering a large range of regions and both
522 monovarietal wines and wines composed of different grape varieties (Cabernet Sauvignon, Merlot, Pinot Noir, Syrah, Grenache, Carignan, Gamay, Mourv6dre). The wines selected for the study included 13 French wines and 1 wine from Italy: Bordeaux (BD): Chfiteau Branaire 1998 (BD/B), Plaisir de Haut-M6doc (BD/HM), Chfiteau la Grave ~ Pomerol 1999 (BD/P), Chfiteau Roche Guitard 2000 (BD/RG); Bourgogne (BG): Aloxe Corton 1998 (BG/AC), La Chance au Roy 2000 (BG/CR); Rh6ne (R): Domaine de Panisse 2001 (R/P), Crozes Hermitage 2002 (R/CH), Les Truffiers 2001 (R/LT); Roussillon (RS): Domaine de Terre Rouge 2002 (RS/TR); Loire (L): Alison 2002 (L/A); Beaujolais (B J): Beaujolais Delhaize 2002 (B J/B), Morgon 2002 (BJ/M); Emilia-Romagna (ER) and from Italy: Villa Giulia 2002 (ER/VG).
2.2. HS-SPME-ChemSensor analysis The hyphenated configuration consisted of a sample preparation autosampler (MultiPurposeSampler| or MPS-2| Gerstel) for headspace-solid phase microextraction (HS-SPME), a 6890/5973 GC-MS system (Agilent Technologies) and a workstation with ChemSensor software (Agilent Technologies) and Pirouette@ pattern recognition software (Infometrix). HS-SPME parameters were evaluated and optimised: 10 ml samples of red wine were each diluted to 10% (v/v) ethanol, 2 g of sodium chloride was added in 20 ml vials and incubated for 3 rain at 40 ~ in the thermostatic agitator of the MPS-2| The sorption of wine volatile compounds was performed for 30 min on a polydimethylsiloxane (PDMS) fibre (100 pro) (Supelco). The GC column was continuously held at 250 ~ and helium was used as carrier gas (1 ml/min). The transfer lines were maintained at 280 ~ The total ion current (70 eV) was recorded in the m/z range from 40-230 amu (scan mode) using a solvent delay of 2 min and a run time of 5 min. For each wine sample a total mass spectrum was generated and converted by the ChemSensor software to a composite mass fingerprint, which could be easily imported into the Pirouette| pattern recognition software. Different pattern recognition algorithms were used for on-line data processing of the mass fingerprints (e.g. principal components analysis (PCA), hierarchical cluster analysis (HCA) and soft independent modelling of class analogy (SIMCA)). 2.3. HS-SPME-GC-MS analysis The hyphenated ChemSensor configuration was also used in the GC-MS mode. Therefore, the GC column (HP-PONA 50 m x 0.2 m m x 0.5 pro, Agilent Technologies) was temperature-programmed: 40 ~ for 5 min, from 40 ~ to 200 ~ (5 ~ from 200 ~ to 248 ~ (8 ~ 248 ~ for 5 rain. A solvent delay of 6 rain was used and the total run time was 48 rain. For GC-MS analysis an aliquot of the internal standard nonane was added to the vials. Each wine was analysed in triplicate. Semi-quantitative determinations of the wine volatiles were obtained by relating the peak areas of the volatiles to the peak area of the internal standard. Principal components analysis (PCA) was performed on the semi-quantitative GC-MS data using The Unscrambler@ (Camo).
523
3. RESULTS The GC-MS data f o r the red wines were analysed by PCA and Figure 1 shows the relationships between the 14 red wines the 70 volatiles identified in these products. For clarity reasons mean values of the 3 replicates for each wine were used in the PCA. Bordeaux and Bourgogne wines had negative PC1 scores and PC2 differentiated both type of wines. Bordeaux wines were characterised by higher levels of oak lactones, vitispirane and phenolic components (4-ethylphenol, 4-ethyl-2-methoxyphenol and 2methoxy-4-propylphenol). Characterised by positive score-values on PC1 were Rh6ne/Roussillon and Loire/Beaujolais/Emilia Romagna. Rh6ne wines. These wines were characterised by high levels of flowery terpene alcohols (linalool and geraniol), while the Beaujolais wines had high levels of acetate esters.
PC2
1.0-
(7%)
2-methylpropyl hexanoate
~
9 RIP
1 -hexanool
BG/CR 9 BG/AC 0.5--
ethyl 2-hexenoateo
9
.
linalool
~
1-nonanol o citronellol o ethyl propanoate o
.e l~rr geranio~eR .....
oBHT
o
i~
o methyl octanoate
01 - o c t a n o l
o ethyl cinnamate oethylbutano t benzyl alcohol o ^ et~[y~,hexanoate butyl benzoate o2-~eetxha~BrcOP~i~ctanoate 2-methyl-l-propanol o o ..oe~ff~lecanoate ...... ethyl acetate, o e ~~ , ~met - ~Y~ro r o a nn o~a% o ~ e oo ois~176 te 9 u ~'l~re(~,~ furfu~yl formate o ,~%1~r4##~,~l'-~-,[,(~4r,~ne-t-hYIPh e n Yl) met h Yl] be n ze n e 4 - et h y I- 2 -cm~_tohaOkX~)n~l~ 2-r~ethyll ,ropyl - "6'm"Xe ' {~-~'~'~,~,~lr~e~oate.nethyl decanoate R nll~ diethyl butanedioate o ^acetate o limone'l'f6~'~'~' ~ hylbutyl decanoate ~ ~ - -9 4I hen ~ Io " .... trans_oap~tlhaY~ne p-cymen, ~ e~v~a~ ~e ,= emyl ~-[memy=m=o&pr0panoateo . ~trans-6-cramffscenone w =r~ r.~ Ar=r-t,,'~" vit=spiraneu 01 t~.c~ e o t e e I a etate R n l l = l i ~Jl [~ g l l ~ ~ ethyl 02-methylpropanoat~ o b e n zal deh'.Y.t~0~,~.,~~ [ ~ y ~ / ~ t ~ - decanoate li~'ll IIII o3methylg[J-ty~' ~an=' ' % a ~e~ JIM th i -met yibuta o te o ~-~. ~o.xy-I~-p.rop~/~p~e.nol o ethyl 2-methyl~utyl .I;)dla.ngaLoate~ z- memw- ] -.0utanot o oethyl 9-decenoate ethyl 3-methvlbutyl but~nedloat.e o 9 ethyl. 3-methylbutanoate o ~-terpmolene o3-methylbutyl acetate 9 ~ ~ J r ~ i ~ ~o..,, 3 4-dichlorobenzenamine v e ,, 0 3-methvlbutvl octanoate ~ 1 ! 1 w ~ 1 I-1[ I ]III-I ' 2-phenylethanol V2-phenyle'thyl &cetate 9 ~-~ m/i-J ll~W' I I ~ ' E l l , , J 3-methyl-l-butanol 0 ,,^ ~ hy!b.utyl o.c.tanoate 1[~4,1/1~ ,,~'-memylDutyl acetate
R/LT
0 ' ,
-0.5--
_ --
-I .0-
.
. . . .
-8.a . . . .
.
8 PC1 (30%)
.
.
.
.
.
.
0'.5
I
1.0
Figure 1. PCA biplot of the volatiles of the 14 red wines from different regions. For explanations on the sample abbreviations see section 2.1. The GC-MS-PCA classification was in good correspondence with principal components analysis of the MS fingerprinting data. In Table 1 the SIMCA interclass distances of the red wines obtained by ChemSensor analysis are presented. Analogies in mass fingerprims are reflected by low interclass distances. An interclass distance lower than 4 is generally an indication for similarity. A good agreement between the MS fingerprinting and the GC-MS profiling approach was obtained.
524
Table 1. Interclass distances (SIMCA) between the 14 red wines obtained from ChemSensor analysis. For full names of the abbreviations used, see section 2.1.
BD/B
BD/ B
BD/ HM
BD/ P
BD/ RG
BG/ AC
BG/ CR
R/ P
R/ CH
R/ LT
RS/ TR
L/ A
BJ/ M
ER/ VG
-
4.50
1.60
3.03
6.80 6.19 12.57 9.57 11.65 11.41 9.30 9.37 10.76 16.25
BD/HM
4.50
-
3.21
3.56
6.73 6.15 8.66 6.07 8.35
8.10 6.24 6.69 8.47
BD/P
1.60
3.21
-
3.04
5.28 4.88 9.78 6.51 8.44
8.34 6.64 6.35 7.59 10.55
BD/RG
3.03
3.56
3.04
-
BG/AC
6.80
6.73
5.28
3.53
BG/CR
6.19
6.15
4.88
3 . 4 1 2.62
R/P
12.57 8.66 9.78
6.95
R/CH
9.57
5 . 0 1 4.67 2.91 3.59
R/LT
11.65 8.35
8.44
6.49
5.55 3.28 2.75 2.34
RS/TR
1 1 . 4 1 8.10
8.34
6.64
5.81 4.40 4.48 3.39 3.32
6.07
6.51
3.53 3.41 6.95 5.01 6.49 6.64 5.04 5.29 6.60 -
2.62 6.39 4.67 5.55 -
6.39 4.76
4.76 2.91 3.28 -
3.59 2.75 -
4.48 4.25 4.54 4.84
2.34 3.39 -
9.99 9.71
5 . 8 1 5.01 5.05 6.16 13.93 4.40 3.37 2.72 3.69 1.97 2.31 2.65
3.32 2.46 2.06 2.83 -
1.97 2.46 2.47
2.47 3.89 4.86
L/A
9.30
6.24
6.64
5.04
5 . 0 1 3.37 4.25
BJ/M
9.37
6.69
6.35
5.29
5.05 2.72 4.54 2.31 2.06
3.89
BJ/B
10.76 8.47
7.59
6.60
6.16 3.69 4.84 2.65 2.83
4.86 2.18 1.08
ER/VG
16.25 9.99 10.55 9.71 13.93 7.78 8.66 6.46 3.72
i
BJ/ B
1.34
1.34 2.18 -
7.78 8.66 6.46 3.72 7.22 3.61
1.08
2.55
-
2.98
7.22 3.61 2.55 2.98
-
4. D I S C U S S I O N A N D C O N C L U S I O N This work resulted in a good accordance between both G C - M S profiling and MS fingerprinting. The a u t o m a t e d C h e m S e n s o r approach appears to be promising for fast objective quality evaluation o f wine. Future publications dealing with different wine types, e.g. with Cabernet Sauvignon or Merlot as d o m i n a t i n g grape, will illustrate the suitability o f the m e t h o d for objective selection of wines as rapid screening m e t h o d for wine importing companies. References 1. J.-L. Le Qu6r6 and P.X. Eti6vant (eds.), Flavour research at the dawn of the 21st century, proceedings of the 10th Weurman flavour research symposium, Paris, France (2003) 572. 2. T. Hofmann, M. Rothe and P. Schieberle (eds.), State of the art in flavour chemistry and biology, proceedings of the 7th Wartburg symposium, Garching, Germany (2005) 98.