Assessment of Trentingrana cheese ageing by proton transfer reaction-mass spectrometry and chemometrics

ARTICLE IN PRESS

International Dairy Journal 17 (2007) 226–234 www.elsevier.com/locate/idairyj

Assessment of Trentingrana cheese ageing by proton transfer reaction-mass spectrometry and chemometrics Eugenio Apreaa,b,, Franco Biasiolia, Flavia Gasperia, Daniela Motta, Federico Marinic, Tilmann D. Ma¨rkb,d a

IASMA Research Center, Agrifood Quality Department, Via E. Mach, 1, 38010 San Michele all’Adige, TN, Italy b Institut fu¨r Ionenphysik, Universita¨t Innsbruck, Technikerstr. 25, A-6020 Innsbruck, Austria c Dipartimento di Chimica, Universita` di Roma ‘‘La Sapienza’’, P.le A. Moro 5, 00185 Roma, Italy d Department of Plasmaphysics, University of Bratislava, SK-84248 Bratislava, Slovak Republic Received 23 March 2005; accepted 20 February 2006

Abstract Proton transfer reaction-mass spectrometry (PTR-MS) data have been analysed by chemometric techniques to monitor cheese ageing by means of on-line direct head-space gas analysis. Twenty cheese loaves of Trentingrana, a trademarked cheese produced in northern Italy, of different origin and ripening degree, were sampled over the whole Trentingrana production area. An increase of the spectral intensity with ripening has been observed for most of the PTR-MS peaks and a univariate analysis identified 16 mass peaks that were significantly different for ripened and young cheeses, respectively. Moreover, the usefulness of different discriminant analyses and class modelling techniques have been investigated. Discriminant Partial Least Squares analysis, while indicating average behaviour and possible outliers, was not able to correctly classify all samples. Soft class modelling performed better and allowed a 100% correct classification. Partial least square calibration predicted the ageing time of each loaf with reasonable accuracy with a maximum crossvalidation error of 3.5 months. r 2006 Elsevier Ltd. All rights reserved. Keywords: Grana cheese; Cheese ripening; Proton transfer reaction-mass spectrometry; Head-space analysis; Chemometry; Volatile organic compounds

1. Introduction Grana Padano is one of the best known Italian hard cheeses and its long tradition and typicality is well recognised and certified by a protect denomination of origin (EC, 1996). Within the consortium of Grana Padano, a more specific geographic identification with its own trademark is recognised as ‘‘Trentingrana’’ (TG) referring to loaves produced in the territory of Trento province (northern Italy). TG, belonging to the family of Grana Padano, is produced with partly skimmed, raw cows’ milk acidified with a mixture of Lactobacillus and coagulated by the addition of bovine liquid rennet but, having a more restrictive production protocol, the use of lysozyme is not allowed. Further difference is the Corresponding author. Tel.: +39 0461 615388; fax: +39 0461 650956.

E-mail address: [email protected] (E. Aprea). 0958-6946/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.idairyj.2006.02.008

prohibition to use silage for the cows’ feeding (Carini & Lodi, 1997). Even if the production of TG is relatively small (about 4000 ton year
1; consortia personal communication) when compared with that of the direct competitors (about 125 000 ton year
1 for Grana Padano; and about 113 000 ton year
1 for Parmigiano; http://www.granapadano.com, www.crpa.it), it is relevant for the local economy. Cheese flavour compounds result from the action of microorganisms and enzymes on the carbohydrates, fats and proteins of the milk and curd and the principal biochemical pathways for their formation have been reported by many authors (Kristoffersen, 1973; Manning, 1979; Engels & Visser, 1994; Fox, Singh, & McSweeney, 1995; Fox & Wallace, 1997; McSweeney, 1997) and reviewed by McSweeney and Sousa (2000). The volatile components of dairy products have been widely studied (Aston & Dulley, 1982; Law, 1984; Manning, Ridout, &

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

Price, 1984; Yvon & Rijnen, 2001; Adda, 1986; Liu, Holland, & Crow, 2004) and more than 600 volatile compounds have been identified in cheese (Maarse & Visscher, 1989; Stratton, Hutkins, & Taylor, 1991). In particular, the volatile organic compounds (VOCs) present in Parmigiano Reggiano and Grana Padano have been extensively studied by chromatographic techniques (Bottazzi & Battistotti, 1974; Dumont, Roger, & Adda, 1974; Meinhart & Schreier, 1986; Barbieri et al., 1994; Moio & Addeo, 1998). An alternative approach to the expensive and time-consuming chromatographic techniques is the use of direct headspace-mass spectrometry for analysis of VOCs (Marcos Lorenzo, Perez Pavon, Fernandez Laespada, Garcy´a Pinto, & Moreno Corsero, 2002). Along this line we characterised TG cheese by means of a direct headspace analysis using proton transfer reactionmass spectrometry (PTR-MS). As indicated by Boscaini, Van Ruth, Biasioli, Gasperi, and Ma¨rk (2003) when analysing similar cheese samples, the choice of PTR-MS has several advantages because of a number of interesting characteristics: (i) PTR-MS is fast, time-dependent variations of the headspace can be monitored with a subseconds time-resolution and the headspace of a sample can be measured in a few minutes; (ii) samples are not subjected to any previous treatments thus reducing the risk of artefacts; (iii) mass spectral intensity can be transformed into absolute headspace concentrations in principle without any calibration with external standards; (iv) PTR-MS has a detection limit in the ppt range (Lindinger, Hansel, & Jordan, 1998) and a high dynamic range of several orders of magnitude. Probably, the main drawback of a direct PTR-MS analysis is that, lacking of a chromatographic separation and of an exhaustive database on pure compounds fragmentation, the interpretation of PTR-MS spectra in the case of complex mixtures, remains a difficult issue. However, for the majority of the compounds reported in the paper the fragmentation pattern acquired in standard condition (the same we used) can be found in

227

Buhr, van Ruth, and Delahunty (2002) and a further interpretation can be based on the work of Boscaini et al. (2003) on similar samples. Tentative peak attribution has been assisted by peaks correlation, isotopic ratios, and by further comparison between headspace evolution of cheese and pure compounds. The possibility to follow on line the decrease of compounds in the headspace of liquid solutions allows sometimes the separation of isobars (Karl, Yeretzian, Jordan, & Lindinger, 2003; Pollien, Jordan, Lindinger, & Yeretzian, 2003; Yeretzian, Jordan, & Lindinger, 2003). This has been used, e.g., to estimate acetaldehyde in cheese based on the intensity of the signal at m=z ¼ 45 (Aprea et al., 2003). We studied Trentingrana samples of different ages from different origins in order to demonstrate the possibility to characterize certain features of TG cheese, i.e., in particular for age prediction by PTR-MS analysis. Different chemometric methods have been applied to the PTR-MS fingerprint data and their usefulness for age prediction has been evaluated. In a recent work (Biasioli et al., 2006) correlation among PTR-MS spectral fingerprinting and odour and flavour sensory profile of TG has been investigated. 2. Materials and methods 2.1. Cheese samples Twenty cheese loaves of TG, aged between 8 and 28 months, were obtained from the main cheese-factories located in Trentino province (Italy) and certified by the ‘‘Trentingrana’’ consortium (the actual sampling took a period of 3 months). Each cheese factory provided two samples: a ‘‘ripened’’ one (not less than 18 months) and a ‘‘young’’ one (less than 15 months) (Table 1). All the cheese factories followed the same protocol (Battistotti & Corradini, 1993) for the cheese production, the only relevant declared difference was that for two factories

Table 1 The factories of the Trentingrana consortium where the analysed cheeses have been sampled Cheese making factory code

Age of the loaves (months) and production period Young

Ab Bb C D E F G H I J

(9) (11) (8) (11) (12) (11) (14) (11) (9) (12)

Production: loaves year
1

Management systema

6292 6292 5226 5260 6480 6622 9182 3218 8750 6745

M,T M,T T M,T M,T T T T M M

Ripened Aug‘01 Sep‘01 Sep‘01 Jul‘01 Jul‘01 Jul‘01 May‘01 Aug‘01 Aug‘01 Jul‘01

(21) (22) (25) (22) (23) (28) (23) (23) (19) (18)

Aug‘00 Oct‘00 May‘00 Aug‘00 Aug‘00 Mar‘00 Aug‘00 Aug‘00 Oct‘00 Jan‘01

T ¼ traditional management system: small farms (o20 cows), milk production o25 L day
1cow
1, feeding with local hay and low level of concentrates. M ¼ modern management system: medium (20–40 cows) or large (440 cows) farms, milk production of 25–40 L day
1cow
1, feeding with high level of concentrates. b A and B refers to the same cheese factory that provided four loaves in two different sampling. a

ARTICLE IN PRESS 228

E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

(E, J) the milk was cooled down and kept at 12–14 1C overnight the day before the cheese making. From a piece of cheese of about 5 kg (height 10.5–11.5 cm, radius 20–24 cm), corresponding to about a 1/8 of a loaf we removed the outer layer (at least 4 cm). The sample for the analysis was a slice of about 2.5  4  12 cm from the remaining inner part of the loaf. This slice was grated and well mixed to reduce the variability of the samples due to the differences in the different parts of the cheese loaf. An amount of 2.5 g of this grated cheese was deposited in glass bottles of 120 mL (Supelco, Bellefonte, USA) sealed with a cap provided with teflon/silicone septum and stored in a refrigerator at 4 1C for no longer than 7 h without any further treatment. After the bottles closure they were not opened and the headspace was sampled through a needle piercing the septum. For each cheese three vials were prepared. 2.2. PTR-MS The PTR-MS technique has been extensively discussed in a series of review papers (Hansel et al., 1995; Lindinger et al., 1998), here we will just recall some general aspects. Introduced by Lindinger and co-worker in 1993 (Lindinger, Hirber, & Paretzke, 1993; Lagg, Taucher, Hansel, & Lindinger, 1994), it is based on a novel design of the chemical ionisation method (Munson & Field, 1966). The sample gas is continuously introduced into a drift tube where it is mixed with H3O+ ions formed in a hollow cathode ion source. Volatile compounds that have proton affinities higher than water (proton affinity of H2O: 166.5 kcal mol
1) are ionised by proton transfer from H3O+, mass analysed in a quadrupole mass spectrometer and eventually detected as ion counts per second (cps) by a secondary electron multiplier. A commercial PTR-MS apparatus (Ionicon GmbH, Innsbruck, Austria) was used and the mass spectrometric data were collected over a mass range from m/z 20 to m/z 259 using a dwell time of 0.2 s per mass (in each cycle a complete mass spectrum up to mass 259 amu is monitored within a time span of 48 s) under drift tube condition of 120–130 Td (Td ¼ Townsend; 1 Td ¼ 10
17 V cm2 mol
1). Each sample was measured for eight cycles and the mean of cycles 3–7 was used for further analysis. Consecutive samples were alternated with blank (empty bottle) whose signal was subtracted from sample spectra. The data collected were converted in ppb according to a procedure given by Lindinger et al. (1998). We assumed the reaction rate constant to be 2  10
9 cm3 s
1 for all compounds, thus introducing a systematic error that is often reasonable (Lindinger et al., 1998). This effect must be considered if comparing our data with results of other techniques but it does not affect the proposed chemometric analyses which require only that the measuring conditions are constant. The problem of signal drifts related to the fingerprinting based-sensor techniques (Kress-Rogers, 1996) should be overcome because the conversion from ion counts into the absolute concentration (ppb) takes into

account the parameters that can change in time and with different apparatuses (Lindinger et al., 1998). Moreover we measured mostly young and ripened cheeses together (randomising the order of samples and replicates) thus an instrumental shift, if present, cannot induce a separation between the investigated classes of cheese. 2.3. Headspace analysis Samples were placed in a water bath at 36.6 1C one hour before and during PTR-MS measurements. The vial was connected via a heated (70 1C) capillary line made of uncoated deactivated fused silica tubing with an inner diameter of 0.25 mm (Supelco, Bellefonte, USA) terminating in a stainless steel needle to the drift tube of the PTRMS (Biasioli, Gasperi, Aprea, Colato et al., 2003). The difference in pressure between the drift chamber (2.0  10
3 bar) and the vial (atmospheric pressure) generates a flux flow of about 10 cm3 min
1 (due to the line impedance), this flux correspond to a velocity of about 3.4 m s
1 resulting in a time of about 0.6 s for the gas to cover the entire length of the transfer line. To avoid a pressure drop during the headspace sampling, the removed gas mixture was replaced with pure nitrogen gas (SOL s.p.a., Italy; purity: 99.999%) through a second stainless steel needle connected to a nitrogen cylinder. The samples were measured with a time interval of 20 min and considering 1 h of conditioning in the water bath, 5–7 h passed between the measure of the first sample and the last one. The measuring order of samples and replicates was randomised to avoid memory effect and nitrogen was flushed, for about 10 min, to clean the transfer line between two consecutive measurements. 2.4. Data analysis PTR-MS is a spectrometric technique, whose output is a high dimensional vector (hundreds of masses) of intercorrelated data (the absence of pre-separation produces interference of different compounds or fragments on the same spectrometric peaks). The analysis of this kind of data is efficiently performed by multivariate analysis in two steps: (i) data compression to reduce the dimensionality of the problem without a significant loss of information and (ii) classification to identify possible interesting structure in the data set. In a first approach to the problem, data compression was performed by means of discriminant Partial Least Square analysis (dPLS) (Kemsley, 1998) that is a restriction of the Partial Least Square technique (Beebe & Kowalsky, 1987), where the additional information (category index) is (i) used to maximise between-class separation and is (ii) useful in finding the variates responsible for group differences. Linear discriminant analysis (LDA) (Duda, Hart, & Stork, 2001) is applied after this compression phase. The LDA method is a pure classification technique, it uses linear boundaries (hyperplanes) to delimit the class regions.

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

In this way, the classification rule can be used for predictions of unknown samples. However, to have an accurate estimation of the predictive ability of the model in the case of unknown samples, a further validation step is needed. In this study, a ‘‘leave-one-out’’ cross-validation approach was used (Stone, 1974; Good, 1999). It should be pointed out that the predictive ability of a classifier could also be affected by data pre-treatment. In particular we noted in previous studies that LDA on PTR-MS spectra performs better if each spectrum is normalised to unit area before any statistical processing (Biasioli, Gasperi, Aprea, Colato et al., 2003; Biasioli, Gasperi, Aprea, Mott et al., 2003). Therefore also in this study the spectra were normalized before data analysis. A different approach to the problem of pattern recognition is the use of class-modelling instead of the use of pure classification techniques. Here we used the UNEQ (UNEQual disperses classes) technique which focuses on one category at a time, defining a boundary to separate a specific class from the rest of the universe. The class space is built according to the Mahalanobis distance as the confidence hyperellipsoid that accounts for 95% of the probability in the case of multivariate normal distribution. This approach seems to be more powerful and more appropriate when dealing with real-world problems. However, its implementation requires a high ratio between the number of samples for each class and the number of experimental variables (more than 3), so a preliminary variable reduction stage is often necessary. In this study the variable reduction has been carried out using a stepwise procedure, based on the value of Wilks’ lambda, which is an inverse multivariate measure of the discriminating ability of a set of variables (Wold et al., 1984). The preliminary data exploration, standard univariate data analysis and analysis of variance (two groups: young and ripened, ten samples per group, three replicates for each sample) were performed with the software package Statistica 5.0 (StatSoft, Inc., Tulsa, OK, USA) on spectra normalised to unit area. Discriminant PLS analysis, as reported by Kemsley (1998) and already tested on PTR-MS spectra (Biasioli, Gasperi, Aprea, Colato et al., 2003) was implemented by means of the software Win-DAS (John Wiley & Sons, Ltd: Chichester, UK, 1998). All the other computations have been performed using the package V-Parvus 2003 (Forina, Lanteri, Armanino, Cerrato-Oliveiros, & Casolino, 2003). 3. Results and discussion 3.1. Descriptive statistics The original data matrix consisted of 60 objects (rows), corresponding to the cheese samples, and 240 variables (columns), corresponding to the headspace concentration of all the measured masses from m/z 20 to m/z 259, data were converted in ppb. We expected a great variability among the samples, because, even if controlled by a single

229

consortium, GT cheese is produced, ripened and stored in small cheese factories scattered in the Trentino province valleys (200 m a.s.l.) and mountains (up to 1000 m a.s.l.). At first glance of the data seems that the spectra of the two groups (‘‘young’’ and ‘‘ripened’’) are very similar but the average intensities, anyway, even if an overall increase of the intensity is the main effect, every sample has its own spectrum with peculiar details. Despite this significant difference between the two groups, on the contrary we did not observe significant differences among replicates nor effects related to the measuring session. From an ANOVA analysis several masses can be identified which are significantly different for the two groups, i.e., 16 masses with a confidence greater than 99.9% (m=z ¼ 37, 43, 45, 47, 61, 63, 65, 71, 89, 91, 103, 107, 109, 115, 117, 119, 145), 24 with a confidence between 99.0% and 99.9%, and 24 with a confidence between 90% and 99%. Here we do not count the peaks which intensity is that expected for isotopes of other masses. We can conclude that the two groups are statistically different and that many peaks have a significantly higher level in the case of ripened cheeses. For the most significant masses a tentative identification based on literature data (Boscaini et al., 2003; Bottazzi & Battistotti, 1974; Dumont et al., 1974; Meinhart & Schreier, 1986; Barbieri et al., 1994; Moio & Addeo, 1998) and pattern fragmentation of pure standards (Buhr et al., 2002) is given in Table 2. In general, an increase of the peak intensity for the ripened cheeses is evident. Moreover, it seems that in young samples the headspace composition is more homogenous, while during the ripening the sample of each cheese factory follows a different path and thus differences between samples are increasing. For example, the signals of the masses related to esters (m/z 89, 103, 117, 145) increase with ripening for the majority of the samples. Esters are recognised as common constituents of cheese flavour. Esterification reactions occur between short-chain and medium-chain fatty acids and primary and secondary alcohols derived from lactose fermentation or from amino acid catabolism during cheese ripening (Curioni & Bosset, 2002). High concentration of ethyl esters in ‘‘Parmesan’’ cheese (Dumont et al., 1974; Meinhart & Schreier, 1986) and in ‘‘Grana Padano’’ cheese (Moio & Addeo, 1998) has been reported previously. In particular, ethyl hexanoate is the major ester present in both cheeses followed in abundance by the ethyl esters of butanoic, octanoic and decanoic acids (Meinhart & Schreier, 1986; Moio & Addeo, 1998). The most intense peaks for these compounds are expected at m/z 145 for ethyl hexanoate, m/z 89 and 117 for ethyl butanoate, m/z 173 for ethyl octanoate and m/z 201 for ethyl decanoate (Buhr et al., 2002). As an example for the tentative identification of a PTRMS peak we report the case of mass 45. The most intense signal for ripened cheeses was recorded at m/z 45 with an average value of 17.5 ppmv (CV% 92) showing an elevated broadening of the sample, while in young cheeses a much lower average value of about 1.0 ppmv (CV% 61) was

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

230

Table 2 Masses, with tentative identification, showing significant different intensities (po0.001, on data normalised to unit area) in young and ripened cheesesa m/z

Class

Mean (ppbv) 43

CV (%)

2376

35

61

Ripened Young Ripened Young Ripened Young

3898 1038 17503 1152 2255 1197

43 63 94 64 46 45

63

Ripened Young

1759 68

43 40

71

Ripened Young Ripened Young

913 230 449 120

95 56 49 46

89

Ripened Young

299 379

66 47

103

Ripened Young Ripened Young

1120 13 32 18

75 21 49 52

107

Ripened Young

36 36

53 46

109

Ripened Young

109 6

77 30

Ripened Young Ripened Young Ripened Young Ripened Young Ripened

7 52 95 39 240 4 7 4 29

36 50 51 57 80 44 46 124 122

45 47

65

91

115 117 119 145

Young

Tentative identificationb

All samples

Fragment common to several compounds Acetaldehyde Ethanol Acetic acid; acetyl esters fragment Acetaldehyde-water clusterc Ethanol-water cluster C-5 alcohols; C-4 acids fragment Ethyl acetate; butanoic acid – Methyl butanoate; isovaleric acid Benzaldehyde; ethyl benzene; o,p,m-xylene Trans-2-octenal; dcarvone fragment 2-heptanone Hexanoic acid; C-6 esters 2-butoxyethanol Ethyl hexanoate

a Headspace average concentration (three replicates) expressed in ppbV and coefficient of variation (%) is reported for the selected masses. b Identification made comparing literature data (VCF database) and fragmentation patterns (Buhr et al., 2002). c From the consideration reported in the text and from the data about GC-O reported in Boscaini et al. (2003) the presence of dimethyl sulfide for GT is below the instrumental error.

observed. We attributed the signal recorded at m/z 45 almost exclusively to the presence of protonated acetaldehyde (acetaldehyde molecular weight 44 amu). We estimated the possible contribution to the signal recorded at m/z 45 from fragments of other compounds comparing the spectra of cheese with the fragmentation pattern of pure compounds. For example, the contribution of 3-methyl-

butanal at mass 45 should be negligible. Mass 69 is the main fragment of 3-methyl-butanal (about 73%), if we suppose that mass 69 is only 3-methyl-butanal (is not this the case), its contribution to the intensity recorded at m/z 45 should be below 3%. In addition to the fermentation operated by microorganisms (lactose metabolism) (Ott, Germond, & Chaintreau, 2000), acetaldehyde can also form in threonine degradation, a process that could be of importance during cheese ripening (Engels, Dekker, de Jong, Neeter, & Visser, 1997). An increase of its concentration during ripening was reported for Swiss Emmental (Bosset, Buetikofer, Gauch, & Sieber, 1997) and Pecorino Sardo (Larrayoz, Addis, Gauch, & Bosset, 2001). Moreover, Boscaini et al. (2003) found that the peak at m/z 45 is the most intense in TG headspace and that, being about five and 15 times higher than in Parmigiano Reggiano and Grana Padano, respectively, it could be used as a marker for the discrimination among these cheeses. In this study, having a wider number of samples, we confirm that for ripened GT the signal at m/z 45 is the predominant in the headspace spectra and furthermore it increases with ripening. From the headspace intensity of signal at m/z 45 (ppb), we attributed to acetaldehyde, we estimated the mg of acetaldehyde for kg of cheese (3 mg kg
1; Aprea et al., 2003) and, in spite of headspace concentration seems very high, the results are compatible with those reported for Parmesan cheese (3.4–7.6 mg kg
1; Barbieri et al., 1994). 3.2. Classification and class-modelling The present PTR-MS data matrix (60 samples  240 masses) of the analysed samples was then used to build a classification model to discriminate between ‘‘young’’ and ‘‘ripened’’ cheeses. At first, discriminant partial least squares (dPLS) analysis was carried out, considering each replicate as an individual sample. The first two dPLS scores are plotted in Fig. 1, where the separation between the two classes is evident. Furthermore, the group of ripened cheeses, appears to be split into at least two separate subclasses, one of these groups is formed by the samples HR, ER, JR. The headspace profile of these three samples presents a lower intensity more similar to the spectra of unripe cheeses. The cheese-factories E and J use a mixture of milks from two consecutive milking, in which the milk from the evening milking is kept overnight at 12–14 1C. This could partly be at the origin of the observed differences. Furthermore, the sample JR has only 18 months of ripening so it is reasonable that the headspace profile is closer to the young cheeses. Moreover, Linear Discriminant Analysis based on Mahalanobis distances on the first three PLS scores was applied to the present data. In the modelling phase, all the samples were assigned to the correct group. Then, due to the small number of samples, a ‘‘leave one out’’ cross validation approach was carried out to validate the model. All the validation samples have been correctly classified. Significantly worse results in terms of the overall predictive

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

231

8 6

PLS Score 2 (7.8%)

4 2 0 -2

JR

-4 -6

HR

-8 ER

-12 -10

-8

-6

-4 -2 0 PLS Score 1 (9.7%)

2

4

6

Fig. 1. Plot of the first two dPLS scores for all samples. The two scores separate well the young cheeses (grey symbols) from the ripened (black symbols). For the ripened cheeses there are at least two groups: samples JR, HR, ER form a separate group (the three replicates are encircled).

ability are obtained, if a different validation procedure is followed. In fact, if a 20-fold cross-validation approach, where the three replicates for each samples are kept out at a time, is adopted, a minor success rate of 85% is obtained for the validation set. This indicates that the model performs better when it is built using the maximum variability and also that the misclassification is probably due to the peculiarities of the misclassified samples. Indeed, the nine samples wrongly assigned are the three replicates of HR, ER, and JR, the same that form an own cluster in dPLS plot (Fig. 1). In addition to allowing quality control and product classification, multivariate approaches give also useful hints for chemical analysis. In fact, an inspection of the loadings for the first latent variables can immediately provide an indication of the masses responsible for the observed separation between the classes; it is then possible to try to correlate these masses with chemical compounds. In particular, the analysis of the first loading (Fig. 2), that resembles the shape of PTR-MS data, reveals that the separation along the first PLS component is mainly due to a group of masses related to esters (m=z ¼ 89, 103, 117, 131, 145 and 173). This result confirms our previous observation, reported above, where, as far as the masses related to esters were concerned, higher signals were observed for the ‘‘ripened’’ cheeses with respect to those for the ‘‘younger’’ ones. In this way, it is possible to explain the role of a single compound or a group of compounds in the discrimination. For the alternative classification method considered, the class-modelling technique UNEQ, two variables are sufficient to build the optimal UNEQ model (Fig. 3). As far as the classification and prediction results are involved, a 100% rate of correct assignments has been obtained, even when a 20-fold cross-validation (as the one

Fig. 2. First loading of dPLS analysis. The underlined masses are those related with esters.

Distance to Ripened

-10

95%

Distance to Young Fig. 3. Coomans plot for the UNEQ model: young cheeses are represented as filled circles, while ripened cheeses are shown as filled triangles.

reported above for dPLS, with all the replicates joined into the same cancellation group) was used. So, the classmodelling approach results in a clear improvement of the prediction results. Additionally, the power of class-modelling techniques is that they can provide the experimenter with information about how properly each sample is described by any of the class models. This information can be encoded into two figures of merit, sensitivity and specificity, defined as the non-error rate for each class and the percentage of objects of other classes rejected by the class-model under study respectively. In the present study, while the specificity was 100%, the sensitivity was 97% as indeed one sample for each class (one replicate of JY and one of HR) fell outside the respective 95% class boundary, thus these points being consequently rejected by the class model. The same result can also visualized graphically in the form of a Coomans plot (Fig. 3), whose axes are the

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

distances from each of the two class models, as computed by UNEQ. The figure shows the clear separation between the two classes (all the samples from one class have a significant distance to the other class model), the two outlying observations are represented as points outside the straight lines designating the corresponding class boundaries. 3.3. PLS modelling of the cheese ageing

30 Predicted cross-validated ageing (months)

232

25 20 15 10

A direct modelling of the relation between the massspectral fingerprint and the age of the samples was attempted using a PLS1 approach, coupled to variable reduction based on the relative importance in prediction. The latter procedure is based on the iterative elimination of the variable which contribute less to the model, and has been described elsewhere (Eriksson, Johansson, KettanehWold, & Wold, 1999). A direct PLS1 analysis on all data without any conditioning or variable selection provide a reasonable good estimation of the age of the loaves with a RMSECV of 3 months with only four latent variables. A more accurate analysis can be done as follow. The optimal PLS model was then built using 33 experimental variables (m=z ¼ 37, 41, 43, 45, 46, 47, 49, 55, 57, 59, 63, 64, 65, 69, 71, 73, 74, 75, 79, 81, 83, 87, 90, 91, 95, 99, 101, 115, 117, 121, 135, 139, 145) which appeared to be most correlated to the response. The choice of the model complexity was based on the minimum error in prediction as evaluated by leave-one-out cross-validation, and resulted in the inclusion of 23 latent variables, leading to R2 ¼ 99:8% and Q2 ¼ 99:3%. These results correspond to an average error of less than 0.2 months on the calibration samples and of about 0.4 months on the validation set with a maximum absolute error of 3.5 months (25%). These results can be seen in Fig. 4, where the predicted cross validated y values are compared to the actual values. It is worth mentioning that even if the model with 23 variables is the best, using only five LVs it is also possible to obtain a rather good model (R2 ¼ 92:6 and Q2 ¼ 91:4): this overall difference is reflected in the decrease in the minimum and maximum errors on increasing the number of LVs (from 1.5 and 4.5, respectively, with five LVs to 0.4 and 3.5 with 23) that is compatible with the analysis performed without any preliminary data selection.

differences). The second procedure, based on the UNEQ model showed better performances for the classification, and has the potential advantage that additional groups can be added to the overall scheme at any time, whereas in LDA a completely new model is required if the number of groups changes. The main advantages of using PTR-MS in this kind of study lies in the minimal treatment of the samples (just grinding) thus avoiding possible artefacts, the short time of the analysis and the low cost of a single measurement. Besides classification/discrimination a series of additional chemical information are available. The results of this work indicate that ageing of TG can be monitored, both qualitatively (building a classification model able to discriminate between young and ripened cheeses) and quantitatively (by means of a PLS regression model, with a low prediction error). These results are particularly important, as, in addition to the absence of defects, this variable is one of the main factors defining the economic value of this cheese. Extension of this work to ‘‘Grana Padano’’ and ‘‘Parmigiano Reggiano’’ is planned aiming at finding a rapid tool to characterise and control the production of these important products of the Italian food culture.

4. Conclusion

Acknowledgements

In this paper chemometric procedures implemented on head-space PTR-MS data were developed to characterize the ageing of TG cheese from rapid headspace measurements and to set models for ageing prediction. In particular two procedures were implemented, checked and compared. The first, based on the dPLS procedure for data compression followed by the construction of a LDA model, showed satisfactory results for the classification and has the advantage to be more conservative for data interpretation (loadings of dPLS shows the masses responsible for

Work partially supported by the PAT-CNR project AGRIIND and by the FWF, Wien, Austria.

5 0 0

5

10

15 20 Ageing (months)

25

30

Fig. 4. Predicted cross-validated estimation of cheese ageing vs actual values.

References Adda, J. (1986). Flavour of diary products. In G. G. Birch, & M. G. Lindley (Eds.), Development in food flavours (pp. 151–172). London, UK: Elsevier. Aprea, E., Biasioli, F., Gasperi, F., Mott, D., Boscaini, E., Mayr, D., et al. (2003). Quantificazione di composti volatili in matrici complesse da misure di spazio di testa: acetaldeide in formaggio. In S. Porretta (Ed.),

ARTICLE IN PRESS E. Aprea et al. / International Dairy Journal 17 (2007) 226–234 Ricerca e innovazione nell’industria alimentare, Vol. 6 (pp. 1098–1103). Pinerolo, Italy: Chiriotti. Aston, J. W., & Dulley, J. R. (1982). Cheddar cheese flavour. Australian Journal of Dairy Technology, 37, 59–64. Barbieri, G., Bolzoni, L., Careri, M., Mangia, A., Parolari, G., Spagnoli, S., et al. (1994). Study of the volatile fraction of Parmesan cheese. Journal of the Science of Food and Agriculture, 42, 1170–1176. Battistotti, B., & Corradini, C. (1993). Cheese, chemistry, physics and microbiology. In P.F. Fox (Ed.), Italian cheese (2nd ed.) (pp. 221–233). London, UK: Chapman & Hall. Beebe, K. R., & Kowalsky, B. R. (1987). An introduction to multivariate calibration and analysis. Analytical Chemistry, 59, 1007A–1017A. Biasioli, F., Gasperi, F., Aprea, E., Colato, L., Boscaini, E., & Ma¨rk, T. (2003). Fingerprinting mass spectrometry by PTR-MS: Heat treatment vs. pressure treatments of red orange juice—A case study. International Journal of Mass Spectrometry, 223 & 224, 343–353. Biasioli, F., Gasperi, F., Aprea, E., Endrizzi, I., Framondino, V., Marini, F., et al. (2006). Correlation of PTR-MS spectral fingerprints with sensory characterisation of flavour and odour profile of ‘‘Trentingrana’’ cheese. Food Quality and Preferences, 17(1&2), 63–75. Biasioli, F., Gasperi, F., Aprea, E., Mott, D., Boscaini, E., Mayr, D., et al. (2003). Coupling proton transfer reaction-mass spectrometry with linear discriminant analysis: A case study. Journal of Agricultural and Food Chemistry, 51, 7227–7233. Boscaini, E., Van Ruth, S., Biasioli, F., Gasperi, F., & Ma¨rk, T. D. (2003). Gas chromatography-olfactometry (GC-O) and proton transfer reaction-mass spectrometry (PTR-MS) analysis of the flavor profile of grana padano, parmigiano, reggiano and grana trentino cheeses. Journal of Agricultural and Food Chemistry, 51, 1782–1790. Bosset, J. O., Buetikofer, U., Gauch, R., & Sieber, R. (1997). Reifungsverlauf von in Folien verpacktem Emmentaler Kase, mit und ohne Zusatz von Lactobacillus casei subsp. casei II. Gaschromatographische Untersuchung einiger flu¨chtiger, neutraler Verbindungen mit Hilfe einer dynamischen Dampfraumanalyse. Lebensmittel-Wissenschaft & Technologie, 30, 464–479. Bottazzi, V., & Battistotti, B. (1974). Head space chromatograms of carbonyl compounds in the volatile components of Grana cheese. Scienza e Tecnica Lattiero-Casearia, 25, 11–19. Buhr, K., van Ruth, S., & Delahunty, C. (2002). Analysis of volatile flavour compounds by proton transfer reaction-mass spectrometry: Fragmentation patterns and discrimination between isobaric and isomeric compounds. International Journal of Mass Spectrometry, 221(1), 1–7. Carini, S., & Lodi, R. (1997). Grana Padano. Consideration of starter culture and lysozyme. Latte, 22, 76–79. Curioni, P. M. G., & Bosset, J. O. (2002). Key odorants in various cheese types as determined by gas chromatography-olfactometry. International Dairy Journal, 12, 959–984. Duda, R. O., Hart, P. E., & Stork, D. G. (2001). Pattern classification (2nd ed.). New York, USA: Wiley. Dumont, J.-P., Roger, S., & Adda, J. (1974). Volatile compounds in whole cheese and grated cheese: Parmesan. Lait, 54, 386–396. EC (1996). Official Journal of the European Union, Regulation no. 1107/ 1996 of 21 June 1996. Engels, W. J. M., Dekker, R., de Jong, C., Neeter, R., & Visser, S. (1997). A comparative study of volatile compounds in the water-soluble fraction of various types of ripened cheese. International Dairy Journal, 7, 255–263. Engels, W. J. M., & Visser, S. (1994). Isolation and comparative characterization of components that contribute to the flavour of different cheese types. Netherlands Milk and Dairy Journal, 48, 127–140. Eriksson, L., Johansson, E., Kettaneh-Wold, N., & Wold, S. (1999). Introduction to multi and megavariate data analysis using projection methods (PCA & PLS). Umea˚, Sweden: Umetrics AB. Forina, M., Lanteri, S., Armanino, C., Cerrato-Oliveiros, C., & Casolino, C. (2003). V-Parvus 2003: An extendable package of programs for data explorative analysis, classification and regression analysis. Department

233

of Chimica e Tecnologie Farmaceutiche e Alimentari, University of Genova, Genova, Italy (free download at http://http.parvus.unige.it). Fox, P. F., Singh, T. K., & McSweeney, P. L. H. (1995). Biogenesis of flavour compounds in cheese. In E. L. Malin, & M. H. Tunick (Eds.), Chemistry of structure/function relationships in cheese (pp. 59–98). London, UK: Plenum Press. Fox, P. F., & Wallace, J. M. (1997). Formation of flavour compounds. Advances in Applied Microbiology, 45, 17–85. Good, I. (1999). Resampling methods: A practical guide to data analysis. Boston, MA, USA: Birkha¨user. Hansel, A., Jordan, A., Holzinger, R., Prazeller, P., Vogel, W., & Lindinger, W. (1995). Proton transfer reaction mass spectrometry: Online trace gas analysis at the ppb level. International Journal of Mass Spectrometry and Ion Processes, 149/150, 609–619. Karl, T., Yeretzian, C., Jordan, A., & Lindinger, W. (2003). Dynamic measurements of partition coefficients using proton-transfer-reaction mass spectrometry (PTR-MS). International Journal of Mass Spectrometry, 223–224, 383–395. Kemsley, E. K. (1998). Discriminant analysis and class modelling of spectroscopic data. Chichester, UK: Wiley. Kress-Rogers, E. (1996). Biosensors and electronic noses for practical applications. In E. Kress-Rogers (Ed.), Biosensors and electronic noses: medicine, food, and the environment. Frankfurt, Germany: CRC Press. Kristoffersen, T. (1973). Biogenesis of cheese flavor. Journal of Agricultural and Food Chemistry, 21, 573–575. Lagg, A., Taucher, J., Hansel, A., & Lindinger, W. (1994). Application of proton transfer reaction to gas analysis. International Journal of Mass Spectrometry and Ion Processes, 134, 55–56. Larrayoz, P., Addis, M., Gauch, R., & Bosset, J. O. (2001). Comparison of dynamic headspace and simultaneous distillation extraction techniques used for the analysis of the volatile components in three European PDO ewes’ milk cheeses. International Dairy Journal, 11, 911–926. Law, B. A. (1984). Flavour developments in cheeses. In F. L. Davies, & B. A. Law (Eds.), Advances in the microbiology and biochemistry of cheese and fermented milk (pp. 197–207). London, UK: Elsevier. Lindinger, W., Hansel, A., & Jordan, A. (1998). On-line monitoring of volatile organic compounds at ppt level by means of proton-transferreaction mass spectrometry (PTR-MS): Medical application, food control and environmental research. International Journal of Mass Spectrometry and Ion Processes, 173, 191–241. Lindinger, W., Hirber, J., & Paretzke, H. (1993). An ion/molecule-reaction mass spectrometer used for on-line trace gas analysis. International Journal of Mass Spectrometry and Ion Processes, 129, 79–88. Liu, S.-Q., Holland, R., & Crow, V. L. (2004). Esters and their biosynthesis in fermented dairy products: A review. International Dairy Journal, 14, 923–945. Maarse, H., & Visscher, C. A. (1989) (6th ed.). Volatile compounds in food. qualitative and quantitative data. Zeist I, The Netherlands: Central Institute for Nutrition and Food Research TNO-CIVO. Manning, D. J. (1979). Chemical production of essential Cheddar cheese flavour. Journal of Dairy Research, 46, 531–537. Manning, D. J., Ridout, E. A., & Price, J. C. (1984). Non-sensory methods for cheese flavour assessment. In F. L. Davies, & B. A. Law (Eds.), Advances in the microbiology and biochemistry of cheese and fermented milk (pp. 229–253). London, UK: Elsevier. Marcos Lorenzo, I., Perez Pavon, J. L., Fernandez Laespada, M. E., Garcy´a Pinto, C., & Moreno Corsero, B. (2002). Detection of adulterants in olive oil by headspace–mass spectrometry. Journal of Chromatography A, 945, 221–230. McSweeney, P. L. H. (1997). The flavour of milk and dairy products: III. Cheese taste. International Journal of Dairy Technology, 50, 123–128. McSweeney, P. L. H., & Sousa, M. J. (2000). Biochemical pathways for the production of flavour compounds in cheese during ripening: A review. Lait, 80, 293–324. Meinhart, E., & Schreier, P. (1986). Study of flavour compounds from Parmigiano-Reggiano cheese. Milchwissenschaft, 41, 689–691. Moio, L., & Addeo, F. (1998). Grana Padano cheese aroma. Journal of Dairy Research, 65, 317–333.

ARTICLE IN PRESS 234

E. Aprea et al. / International Dairy Journal 17 (2007) 226–234

Munson, M. S. B., & Field, F. H. (1966). Chemical ionization mass spectrometry. I. General introduction. Journal of the American Chemical Society, 88, 2621–2630. Ott, A., Germond, J. E., & Chaintreau, A. (2000). Origin of acetaldehyde during milk fermentation using 13C-labeled precursors. Journal of Agricultural and Food Chemistry, 48, 1512–1517. Pollien, P., Jordan, A., Lindinger, W., & Yeretzian, C. (2003). Liquid–air partitioning of volatile compounds in coffee: Dynamic measurements using proton-transfer-reaction mass spectrometry. International Journal of Mass Spectrometry, 228, 69–80. Stone, M. (1974). Cross-validatory choice and assessment of statistical predictions. Journal of the Royal Statistical Society. Series B (Methodological), 36(2), 111–147.

Stratton, J. E., Hutkins, R. W., & Taylor, S. L. (1991). Biogenic amines in cheese and other fermented foods: A review. Journal of Food Protection, 54, 460–470. Wold, S., Albano, C., Dum, W. J., Edlund, U., Esbensen, K., Geladi, P., et al. (1984). Multivariate data analysis in chemistry. In B. R. Kowalski (Ed.), Chemometrics, mathematics and statistics in chemistry. Dordrecht, Holland: Riedel Publishing. Yeretzian, C., Jordan, A., & Lindinger, W. (2003). Analysing the headspace of coffee by proton-transfer-reaction mass-spectrometry. International Journal of Mass Spectrometry, 223&224, 115–139. Yvon, M., & Rijnen, L. (2001). Cheese flavour formation by amino acid catabolism. International Dairy Journal, 11, 185–201.