Potential and Limitation On Authentication Food

Food Research International 60 (2014) 189204
Contents lists available at ScienceDirect
Food Research International

journal homepage: www.elsevier.com/locate/foodres
Potential and limitations of non-targeted ngerprinting for

authentication of food in ofcial control
S. Esslinger, J. Riedl, C. Fauhl-Hassek
BfR Federal Institute for Risk Assessment, Department of Safety in the Food Chain, Max-Dohrn-Str. 8-10, D-10589 Berlin, Germany
a r t i c l e
i n f o
Article history:
Received 8 August 2013
Received in revised form 26 August 2013
Accepted 7 October 2013
Available online 14 October 2013
Keywords:
Authenticity
Adulteration
Chemometrics
Honey
Olive oil
Wine
a b s t r a c t
The investigation of the so-called food ngerprints provides high potential with regard to the characterization
and identity verication of food. Therefore, this kind of non-targeted analysis obtained increasingly importance
during the recent years. These applications are usually based on spectroscopic and spectrometric data providing
the capability for a comprehensive characterization of the investigated matrices. The subsequent statistical
multivariate data analysis enables a general identication of many deviations from the expected product
composition. Besides the classical tests of authenticity of foods, a comprehensive analysis that also allows the
detection of hazardous or safety-relevant manipulations and violations of the respective laws e.g. with regard
to non-authorized food additives or a prohibited use of technological processes is of urgent need in food control.
In the literature, several approaches are already pursuing the non-targeted observation of abnormalities in
various foods covering a broad variety of analytical methods. This review highlights a current overview of the
applicability of this approach using classic spectroscopic as well as spectrometric analytical techniques on the
basis of examples of the three most investigated food matrices: honey, olive oil and wine. Furthermore,
difculties as well as challenges regarding the use of food ngerprinting in ofcial food control are discussed.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
The identity and authenticity of products are current topics in food
and feed science for both sides: on the one hand consumer protection
authorities and on the other hand producers and dealers. Since the
beginning of food trade, incidents concerning adulterations of relevant
products are well known. However, detection of these adulterations
provides a great challenge to the analytical chemists concerning its
identication because of increasing product diversity and the continuous
development of new production technologies.
The term authentication (authenticity testing) used in food control
describes the conrmation of all requirements regarding the legal
product description or the detection of fraudulent statements (this
denition is based on (Gary & Ebeler, 2011; Lees, 2003; MoniQA,
2013). Particularly in view of:
(i) the substitution by cheaper but similar ingredients,
(ii) extension of food using adulterant (e.g. water, starch including
exogenous material) or blending and/or undeclared processes
(e.g. irradiation, extraction),
(iii) the origin, e.g. geographic, species or method of production.
The classical authenticity assessment of food is usually based on
the analysis of specic marker compounds, which are indicative
for a certain property of the product, e.g. shikimic acid in wine or
Corresponding author. Tel.: +49 3018412 3393.
E-mail address: carsten.fauhl-hassek@bfr.bund.de (C. Fauhl-Hassek).
0963-9969/$ see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodres.2013.10.015
hydroxymethylfurfural in honey. The comparison of an actual measurement value with a control limit, which is in authenticity testing
often a so-called experience value for a certain parameter, is a common
approach of the control process. This part of the assessment is always
based on whether the analytical parameter determined, considering
also the measurement uncertainty, is violating the established limit or
if it is still in compliance with it. Therefore, state-of-the-art of the ofcial
control practice is characterized by its possibly forensic utilization. Due
to its high demand for reliability e.g. in case of an ofcial objection
standing beyond reasonable doubt, high efforts are put on the
validation of the analytical method to ensure accuracy of the analytical
determination. Moreover conservative data assessment and evaluation
is done in practice.
Food Fingerprinting, the non-targeted chemical analysis of food with
subsequent multivariate data analysis is based on the principle of the
so-called metabolomics. This term has been established in the late
1990s, more specically, Oliver et al. used the expression metabolome
in the year 1998 for the rst time (Oliver, Winson, Kell, & Baganz,
1998). By denition, metabolomics describes the scientic study of
small molecules, the metabolites, of a biological system based on
comprehensive chemical analysis (omics technologies) with the aim to
detect as many substances as possible (Roessner, Nahid, Chapman,
Hunter, & Bellgrad, 2011). Since both, origin and current focus of
metabolomics are in the eld of pharmacy and toxicology, the analysis
of food using these techniques becomes more and more important e.g.
in food and feed science. In this eld, a distinction was made between
the concepts of food ngerprinting and food proling in accordance to
190
S. Esslinger et al. / Food Research International 60 (2014) 189204
the corresponding denitions in metabolomics (Table 1, Koek, Jellema,

van der Greef, Tas, & Hankemeier, 2011). Food proling focuses on the
analysis of a group of metabolic products in combination with a certain
metabolic pathway or a class of compounds (multi-component analysis).
This strategy is based on the prior knowledge of the analyst concerning
the respective present biological system and, hence, is rather performed
by targeted analysis and optional subsequent multivariate data analysis.
In contrast, food ngerprinting techniques do not deal with the identication of all metabolites, but on the recognition of patterns, the socalled ngerprints of the matrix (Antignac et al., 2011). The genetic
background of agricultural commodities and various environmental or
other external inuences affect the ngerprint of food matrices dramatically. For example, wine of the same variety was certainly subject to
different growing and production conditions depending on the vineyard
location, the climate and the applied oenological practices, which to the
end also affect the respective ngerprints. Those manifold inuences on
the food ngerprint make a proof of product identity challenging. The
origin and certain oenological practices could cause specically
different food ngerprints, which might then result in decision-relevant
differences after mathematical analysis. It is therefore important to
analyze the acquired data sets carefully and perform statistics considering
the available meta-data.
Food ngerprinting approaches are typically based on a highthroughput screening of samples (if necessary after a (simple) sample
preparation) with the purpose of a differentiation or classication of
samples. After identication and mapping of the patterns to individual
food matrices, the aim is to differentiate various food ngerprints in
terms of e.g. their botanical or geographical origin on the one hand or
for instance with respect to possible adulterations on the other hand.
The potential investigation of multiple objectives with only one
analytical method is thereby a clear advantage of the non-targeted
food ngerprinting over the classical targeted approaches.
The application of non-targeted food analysis in combination with
the subsequent use of statistical methods to test for identity and
authenticity is a growing eld of food science. Lai et al. was among the
rst in 1994 who carried out investigations in the eld of food ngerprinting (Lai, Kemsley, & Wilson, 1994). In this study the researchers
tested edible oils for authenticity using a Fourier transform infrared
spectrometer (FT-IR) in conjunction with two multivariate statistical
methods, principal component analysis (PCA) and the discriminant
analysis (DA). The authors used the spectral data to successfully
differentiate several edible oils according to their plant species
(e.g. grape seed, groundnut, corn, rapeseed or walnut) and in the case
of olive oils their manufacturing process (cold pressed or rened). Since
that time, the number of institutions dealing with food ngerprinting
and the number of related publications are growing exponentially.
Besides ngerprinting and proling, further terminologies are
mentioned in literature referring to the eld of food authentication.
The most relevant denitions and principles are described in Table 1.
Recently, foodomics has been adopted for the methodology, which
studies the food and nutrition domains through the applications of
advanced omics technologies (Cifuentes, 2009; Herrero, Garca-Caas,
Simo, & Cifuentes, 2010). Hence, food ngerprinting as well as food
proling might be considered as part of foodomics. So far, the term
foodomics has not yet been used widely in science. Food forensics represents a term, not clearly dened in literature. This approach comprises
all techniques to authenticate food (e.g. by stable isotope ratio analysis or
methods based on DNA analysis, proteomics, metabolomics) and might
therefore be used as a synonym for food authentication (Primrose,
Woolfe, & Rollinson, 2010; Teletchea, Maudet, & Hnni, 2005; Woolfe &
Primrose, 2004). However, besides the use by various companies
providing services in authentication or the specic identication of
exogenous particles in food, this term is not established in the scientic
community.
The most commonly used methods in the eld of food ngerprinting
are based on spectroscopic data, for example, generated by using nuclear
magnetic resonance (NMR)-, near-infrared (NIR)- or FT-IR spectroscopy.
These techniques offer the possibility to analyze relatively small amounts
Table 1
Key terms and description of food characterization techniques adapted for authentication issues.
Authentication
approach
Principle
Aim and advantage/disadvantage
Classic targeted
approaches
Bottom-up-approach
Selective sample preparation
Targeted analysis of single
compounds or group of compounds
Qualitative and/or quantitative
Univariate data analysis
Aim: e.g. authentication/characterization of food (Fauhl, 2006; Koek et al., 2011)

Advantage/Disadvantage
High sensitivity, high selectivity
Simple data evaluation
Time consuming (extensive sample
preparation, multiple analysis)
Only known compounds detectable
Aim: e.g. authentication/characterization of food (Antignac et al., 2011; Cevallos-Cevallos et al., 2009; Koek et al., 2011)
Advantage/disadvantage
High sensitivity, high selectivity
Extensive sample preparation
Need of compound data bases for
identication (spectra comparison)
Need of sample databases for authentication
(multivariate modeling)
Top-down-approach
Selective or unselective sample
preparation
Targeted analysis of a group of
compounds or
Non-targeted analysis and
subsequent identication of
compounds
Qualitative and/or quantitative
Multivariate data analysis
Food
Top-down-approach
ngerprinting Unselective sample preparation
Non-targeted analysis of a spectral
ngerprint
Qualitative and/or semiquantitative
Multivariate data analysis
Food proling
Foodomics
Using omics techniques e.g.

genomics, transcriptomics,
proteomics, and/or metabolomics
Aim: e.g. comprehensive authentication/

characterization of food
Advantage/disadvantage
High-throughput approach
Simple or no sample preparation
Detection of unexpected additives/deviations
possible
Investigation of multiple objectives possible
Need of sample databases for authentication
(multivariate modeling)
Aim: investigation of food and nutrition for e.g.
compound proling, authenticity and/or
biomarker detection
Source
(Antignac et al., 2011; Cevallos-Cevallos et al., 2009; Koek et al., 2011)
(Cifuentes, 2009; Garca-Canas, Sim, Herrero, Ibnez, & Cifuentes,

2012; Herrero, Sim, Garca-Canas, Ibnez, & Cifuentes, 2012)
of sample or its extract in a non-destructive, easy, quick and direct (with

or without minor sample preparation) way. Therefore, the application
of these spectroscopic methods represents a suitable strategy for the
characterization of complex biological systems such as foods, since they
allow a simultaneous determination of a high number of compounds. In
addition, mass spectrometry (MS) hyphenated with chromatographic
separation methods (gas chromatography (GC) or high performance
liquid chromatography (HPLC)) or stand-alone is also widely used.
Herein, the key benet lies in the higher sensitivity, so even the smallest
traces of various compounds might be detected. The resulting data sets
of such non-targeted approaches are usually very complex with two
or more dimensions of often several thousand data points, typically
without assignment of signals to specic substances. For their analysis,
different techniques are pursued, mainly chemometric methods of
pattern recognition or multivariate calibration. Food ngerprinting is
also performed using other analytical techniques, e.g. sensory analysis,
electronic nose, or bioanalytics. However, the focus of this review is the
discussion of the mainly used spectrometric (mass spectrometry stand
alone and/or coupled to chromatographic separation techniques) and
spectroscopic techniques (e.g. IR, NMR or UVvis), whereas approaches
by sensory analysis, electronic nose or bioanalytics (including proteomics
as well as genomics) are not considered here.
The aim of this review is to provide an overview on the current
(last ve years) state-of-the-art strategies in the eld of food ngerprinting and to reect the main challenges of these techniques with
respect to food control. The general characteristic steps of this nontargeted technique and the main chemical analytical procedures applied
in the eld are introduced. Furthermore, applications of the three food
matrices, honey, olive oil and wine most frequently analyzed using food
ngerprinting are discussed. The last part of the discussion focuses on
perspectives in ofcial food control and respective central questions
concerning quality assurance and control.
2. Food ngerprinting
2.1. General approach
All procedures of food ngerprinting are performed in a similar
manner, starting with the denition of the study, including a specication
of sample material(s) and analytical technique(s) which will be

used for the investigations (Fig. 1). Common to all investigations
is then the sampling followed by a more or less intensive sample
preparation, the measurement of the food ngerprint with subsequent
data pre-processing and analysis. Optionally, key compounds might be
determined, which are responsible for a discrimination of the matrices.
This general structure of metabolomics strategies has already
been extensively discussed by several reviews (Antignac et al.,
2011; Cevallos-Cevallos, Reyes-De-Corcuera, Etxeberria, Danyluk,
& Rodrick, 2009; Hendriks et al., 2011; Theodoridis, Gika, Want, &
Wilson, 2012) and is only briey referred to in this review as a whole.
Here, the focus is on the practical application of food ngerprinting
(including the acquisition, pre-processing and statistical analysis
of data) and its use in food control. Thereby, general approaches,
challenges and recent developments in analytical instrumentation are
highlighted as well as hurdles and typical strategies to process the data.
Food ngerprinting is a non-targeted analysis with the intrinsic aim
to detect as many components of matrix as possible. Therefore, selection
of the detection principle including sample preparation, apart from
limitations in availability of the analytical instrumentation, represents
a great challenge to the analyst. The metabolome of a food matrix
consists of a wide range of compound classes with different physical
and chemical properties, which could therefore be difcult to detect
with only one analytical measurement/instrument. To approach this
challenge, sample preparation should be as unspecic as possible to
avoid a loss of signals, accompanied with a possible loss of information.
In many studies this step is mainly performed to enable the accessibility
of the sample matrix to the available analytical equipment and not to
clean-up or enrich matrix components of interest. Thus, for example,
an analytical focus in the application of FT-IR is the analysis of oils,
wines or other liquid or viscous matrices, whose investigation usually
takes place directly without further purication. Somewhat more complicated is the sample preparation designed for solid matrices and/or if
the appropriate task or the analytical method permits a direct investigation. In these cases, the entire sample preparation (incl. extraction and
clean-up) has to be as non-specic as possible to capture as much matrix
components as possible. Alternatively, e.g. Frank et al. generated different
extraction fractions of a matrix (barley) using different solvents (aqueous
and organic). These were analyzed separately by a gas chromatography
Experimental design
Sampling
Sample preparation
- definition of the
objective
- type of matrix
- analytical technique
establishment & sampling

of a respresentative sample
set
e.g.
- milling, homogenization
- extraction, clean-up
- derivatization
- addition of buffer
spectroscopic/
spectrometric
data
191
Aquisition of
chemical fingerprint
by the use of spectroscopic
and/or
spectrometric techniques
Statistical
data analysis
Determination of
key compounds
Data analysis
Data pre-processing
e.g. by the use of databases
by the use of supervised

and/or unsupervised
techniques
e.g.
- bucketing,
- normalization, scaling,
- 1st, 2nd derivative,
- feature selection
Fig. 1. The procedure/strategy of ngerprinting techniques for food analysis.
192
system coupled to a ame ionization detector (GC-FID), followed by

multivariate statistical analysis of the combined analytical data (Frank,
Scholz, Peter, & Engel, 2011).
One of the most widely used analytical techniques for food ngerprinting is the NMR spectroscopy. This technique provides some advantages, which makes it a useful tool in the eld of food ngerprinting
(Le Gall & Colquhoun, 2003). For instance, it is possible to analyze
several matrices directly with little effort in sample preparation.
Furthermore, NMR measurements (under consideration of the stateof-the-art technology) are easily repeatable, reproducible and offer
the possibility of structure determinations of key compounds. On the
other hand, there are the high instrumentation costs and the low
sensitivity compared to other analytical techniques, e.g. FT-IR or MS,
respectively. Most NMR studies are based on 1H measurements instead
of 13C measurements, probably due to the higher sensitivity and shorter
relaxation times. A variety of ngerprint studies, dealing with 1H NMR,
are discussed more closely in chapter 2.2. For this reason, only 13C NMR
applications are focused hereafter. Aursand et al. examined the
feasibility to distinguish salmon (Salmo salar) due to his wild, farmed
and geographical origin using 13C NMR and multivariate statistical
analysis (Aursand et al., 2009). Furthermore, a study by Wei et al. showed
the applicability of 13C NMR spectroscopy in combination with multivariate statistics to distinguish the species and origins of green coffee
bean samples of Arabica and Robusta from several different geographic
regions (Wei et al., 2012). Using classication models, the authors
identied metabolites, e.g. sucrose caffeine and amino acids, which
are responsible for the differentiation. Kbler et al. investigated the
possibility of the identication of pine nuts, which cause the so-called
pine nut syndrome. For this purpose, the applicability of the 1H NMR
was compared with 13C NMR (Kbler et al., 2011). In the result, the
usage of 1H NMR provided more (detailed) information. Therefore,
this technique was judged to be sufcient to conduct a discrimination
of pine nuts.
Another noteworthy special NMR application is the two-dimensional
NMR (2D NMR). This technique is based on the coupling of nuclear
dipoles. In a graphical representation both the abscissa and ordinate
represent frequency-axis, whereas the intensities correspond to a third
dimension. These three dimensional data sets, have to be evaluated
using appropriate statistical models. So far, only very few studies
are known, which investigated the feasibility of 2D NMR using nontargeted approaches to reveal food ngerprints and subsequent
statistical analyses. Lolli et al. studied Italian honey by 2D NMR (1H
13
C) with respect to the botanical origin (Lolli, Bertelli, Plessi, Sabatini,
& Restani, 2008). As a result, samples could be easily distinguished
from each other by PCA. In addition, general discriminant analysis
(GDA) showed a capacity of prediction of more than 92%. These results
showed that 2D NMR coupled with multivariate statistical methods
seems to be an appropriate technique to classify honey samples due
to their botanical origin. Van et al. compared potentials of onedimensional (1D) and 2D NMR applications in metabolic proling
(Van et al., 2007). Of course, the acquisition of 2D data is more timeconsuming, but these data then provide more comprehensive, global
metabolic proles enabling clear group separations due to changes of
additionally detected and resolved low-abundance metabolites in the
proles. Moreover, it was shown that the identication of key components is much easier by using 2D than 1D NMR. As 1D 1H NMR
spectra of complex biological samples typically have high spectral
overlap, which strongly limits the number of metabolites that can be
uniquely identied, it is conceivable that in the future 2D NMR might
become a powerful tool for high-throughput screening in the eld of
food ngerprinting, if the interest is on post-component identication.
Vibrational spectroscopy is based on the transitions between
quantized vibrational energy states of molecules, including analytical
measurement methods such as infrared (IR) and Raman spectroscopy
(Li-Chan, Chalmers, & Grifths, 2010). Mid-infrared (MIR, 400
4000 cm 1) and NIR (400014,000 cm 1) as well as Raman
spectroscopy were shown to be most useful tools in non-targeted

analysis by a large number of scientic publications investigating
their applicability in food ngerprinting. In recent years, FT-IR is of
particular emphasis, due to its advantages in a faster acquisition of
spectra and a nowadays improved signal-to-noise-ratio. RodriguezSaona and Allendorf claried in their review the potential of FT-IR
applications in targeted and non-targeted authentication and detection of adulteration of food (Rodriguez-Saona & Allendorf, 2011). In a
similar literature review Cozzolino presented NIR applications with
subsequent statistical differentiation of food due to their genetic
varieties (e.g. lysine mutants in barley) (Cozzolino, 2011). Recently,
Rohman and Che Man discussed in a literature review the application
of FT-IR in food ngerprinting, in particular, the authentication of
edible oils (extra virgin olive oil, virgin coconut oil and cod liver
oil) in terms of respective adulterations with sunower or sesame
oil (Rohman & Man, 2012a). In conclusion, the authors showed that
FT-IR spectroscopy and subsequent suitable statistics seems to be a
simple and fast powerful method for authentication purposes.
In the context of vibration spectroscopy, a research project sponsored
by the European Commission (EC, Sixth Framework Program) needs
to be mentioned. TRACE (Tracing the origin of food) partially aimed
to develop and validate analytical food ngerprinting procedures
enabling the verication of food origin (TRACE, 2013). Within this
ve year project (starting up in 2005) edible oil, honey as well as
beer samples were chosen to investigate the applicability of vibrational
spectroscopic techniques (FT-IR, NIR, 1H NMR and Raman) concerning a
discrimination of different geographical origin (Caetano et al., 2007;
Donarski, Jones, & Charlton, 2008; Hennessy, Downey & O'Donnell,
2008, 2009; Pierna, Abbas, Dardenne, & Baeten, 2011; Pierna et al.,
2012; Woodcock, Downey & O'Donnell, 2008, 2009). Here, Pierna
et al. compared NIR, MIR and Raman spectroscopy (Pierna et al., 2012).
One hundred and thirty beer samples (Trappist and non-Trappist beers)
were analyzed by the three analytical methods with subsequent statistical discrimination between the two groups. In all cases good discrimination rates were achieved. Because of resulting individual spectroscopic
data sets (NIR, MIR as well as Raman data), individual mathematical
models were used. Therefore, nal results could not be clearly compared
with each other, and so, the most appropriate technique for the
mentioned problem was not identied.
The mass spectrometry based food ngerprinting is an emerging
approach in this eld of research. One of the rst studies using MS
was on differentiation and identication of grapefruit juices fortied
with vanillin in 2001 (Goodner & Rouseff, 2001). The authors analyzed
the samples by an ion-trap mass spectrometer chemical sensor (m/z
50200). The rst part of the statistical data analysis was PCA. It was
used for a better overview of the acquired data set as well as to reduce
the amount of variables to perform subsequent discriminant function
analysis (DFA) in a second step. As a result, all fortied samples (different
concentration levels) were tightly clustered and well separated, showing,
that MS is an appropriate tool for food ngerprinting.
The wide eld of chemical analysis using MS increased in the past 10
to 15years concerning ionization techniques accompanied by faster and
higher resolution detection. Particularly noteworthy in the context of
food ngerprinting approaches is desorption electrospray ionization
(DESI) and direct analysis in real time (DART), two upcoming ionization
techniques. Hajslova et al. successfully demonstrated in a literature
review the diversity of DESI and DART applications for food quality
and safety control as well as for identication and authentication of
food (Hajslova, Cajka, & Vaclavik, 2011). The big advantage of DARTMS is the reduced effort for sample preparation due to direct analysis
of the sample at atmospheric pressure, which was for a long time a
limiting factor in non-targeted MS spectrometry. In addition, several
studies testing the authenticity of olive oil used proton transfer reaction
(PTR) mass spectrometry (Ruiz-Sambls et al., 2012) or secondary
electrospray ionization (SESI) mass spectrometry (Martnez-Lozano
Sinues et al., 2012).
Recently, Garrett et al. and Garcia et al. investigated, for instance, the
differentiation of coffee and whiskey, respectively, in terms of their
botanical origin (Garcia et al., 2013; Garrett et al., 2013). Both research
groups successfully used the Fourier transform ion cyclotron resonance
mass spectrometry (FT-ICR-MS) for analysis, a high-resolution technique
providing masses with high accuracy.
Mass-selective detection coupled to chromatographic separation
systems, such as HPLC or GC is also used in food ngerprinting. Vaclavik
et al. discussed for example the possibility to differentiate red wines due
to their geographical origin using HPLC coupled with a time of ightmass spectrometer (TOF-MS) and subsequent PCA and partial least
squares (PLS) analysis (Vaclavik, Lacina, Hajslova, & Zweigenbaum,
2011). However, the focus at using hyphenated techniques is on the
application of headspace-solid phase micro extraction with subsequent
gas chromatographic separation of the matrix components and mass
selective detection (HS-SPME-GC-MS). Various studies, including Guo
et al. investigated avor proles of various foodstuffs (e.g. apple juice)
to differentiate the matrices in terms of geographical or botanical
origin (Guo, Yue, & Yuan, 2012). By using such techniques, increased
mathematical complexity needs to be handled, e.g. three dimensional
data, retention times and mass drifts. Nevertheless, these approaches
seem suitable for such objectives.
After spectroscopic or spectrometric acquisition of the chemical
ngerprints, data pre-processing is usually necessary and performed.
Instrumental analysis, for instance, may lead to signal shifts during the
measurements for various reasons. As an example, chromatographic
separation of the matrix components prior to mass spectrometry may
result in shifts due to minimal changes in the head pressure affecting
the ow rate and thus retention time of the components. The preprocessing then serves to compensate these differences by means of
chemometric methods. Each analytical method requires its own typical
pre-processing steps and methodologies. For instance, FT-IR applications
often use standard normal variate (SNV) processing as well as calculation
of the rst or second derivative (e.g. SavitzkyGolay) to smooth the
spectra and to emphasize differences between spectra. In NMR, for
example, normalization and baseline correction is applied for noise
reduction. The most widely used mathematical approach to minimize
peak shifts in NMR spectroscopy and to reduce the acquired data size
for multivariate statistics is the so-called binning (or bucketing). This
procedure bases on segmenting a spectrum into small areas (bins/
buckets) and taking the area under the spectrum for each segment.
Preferably, the size of the bins should be large enough to keep a given
peak in its bin despite small spectral shifts across the spectra, but not
too large to include peaks from multiple compounds within a single bin.
However, to realize this compromise becomes more difcult the more
complex the investigated matrix is due to an increasing number of
193
overlapping signals. A major drawback might be the loss of a considerable

amount of information enclosed in the original spectra, which seems less
necessary with increasing computational power. For more details on
specic pre-processing steps the reader is referred to Smolinska et al.
for NMR (Smolinska, Blanchet, Buydens, & Wijmenga, 2012), Katajamaa
and Orei for MS (Katajamaa & Orei, 2007) as well as van den Berg
et al. and to Goodacre et al. for particular denitions (Berg, Hoefsloot,
Westerhuis, Smilde, & Werf, 2006; Goodacre et al., 2007). In contrast to
two-dimensional data (from e.g. FT-IR, NIR, 1D NMR or GC-FID), the
handling of three-dimensional data (from HPLC- or GC-MS as well as
2D NMR) or higher (from e.g. GCxGC-MS) is especially challenging in
terms of data pre-processing. In the case of multiway data sets, a number
of software tools were developed to convert this raw data into lowerdimensional data followed by an alignment procedure. Furthermore,
many providers of analytical instruments respond to the growing
demand for such applications with the development of in-house
programs. A good overview of available software can be found on the
website of Fiehn Lab, which is part of the UCDavis Genome Center
(Fiehn Lab, 2012).
The pre-processing prepares the data for statistical analysis, which is
mainly based on multivariate methods. Such statistical techniques are
typically used to reduce the complex data sets, to predict analytical
parameters using calibration methods or to classify the subjects into
specic groups according to different objectives. In the latter case,
multivariate methods can be divided into unsupervised and supervised
approaches. Unsupervised methods used for pattern recognition within
complex spectroscopic or spectrometric data sets aim to identify
clusters or trends among samples. Thereby, a special feature is that no
prior knowledge of classes or groups is required. In contrast, supervised
classication methods are based on a prior creation of classication
rules using a data set with objects of known class membership (training
set). Afterwards unknown objects (test set) could be classied to one of
the existing classes. Commonly used unsupervised methods in the eld
of food ngerprinting are hierarchical cluster analysis (HCA) and PCA.
Popular supervised methods are linear discriminant analysis (LDA)
and soft independent modeling by class analogy (SIMCA). The mainly
used calibration tool in non-targeted ngerprinting analysis is PLS.
Additionally, other multivariate data analysis approaches are
applied in food ngerprinting, especially in the case of a non-linear
problem. Doing so, e.g. Oliveri et al., Pierna et al., Chen et al., Schievano
et al., Zuh et al. or Imparato et al. used unequal class spaces (UNEQ),
potential function techniques (POTFUN) supported vector machines
(SVM) as well as articial neural networks (ANN) as classication
models (Chen et al., 2012; Imparato, Paolo, Braca, & Lamanna, 2011;
Oliveri, Di Egidio, Woodcock, & Downey, 2011; Pierna et al., 2011;
Schievano, Stocchero, Morelato, Facchin, & Mammi, 2012; Zhu et al.,
Table 2
Classical targeted analytical approaches typically applied for the investigation of the most frequent food authentication issues (geographical and botanical origin as well as adulteration)
specied for the food matrices honey, olive oil and wine.
Objective
Geographical origin
Botanical origin
Adulteration
Honey
Olive oil
Wine
Parameter
Lit.
Parameter
Lit.
Parameter
Lit.
Stable isotope analysis

(13C, 15N, 2H, 34S, 18O)
Mineral content
Trace elements
Pollen analysis (microscopy)
Mineral content
DNA analysis
[14]

(13C, 15O, 2H)
Rare-earth elemental composition
Trace elements
Triglyceride prole
Fatty acid composition
Stigmastadiene
Tocopherol content
Triglyceride prole
Fatty acid composition
Stigmastadiene content
Tocopherol content
[3]

(13C, 15N, 2H, 34S, 18O)
Mineral content
Trace elements
Shikimic acid (Burgundy wines)
Anthocyanic pattern (red wines)
[3]

(13C, 15N, 2H, 34S, 18O)
Mineral content
Volatiles
[13,14]
Stable isotope analysis (13C)

Determination of saccharides
Pollen analysis (microscopy)
[1,2,5,6]
[1,2]
[710]
[710]
[11,12]
[1]: (Anklam, 1998); [2]: (De Alda-Garcilope, Gallego-Pic, Bravo-Yage, Garcinuno-Martnez, & Fernndez-Hernando, 2012); [3]: (Drivelos & Georgiou, 2012); [4]: (Vinci, Preti, Tieri, &
Vieri, 2013); [5]: (Waiblinger et al., 2012); [6]: (Laube et al., 2010); [7]: (European Commission, 2013); [8]: (Christie, 2003); [9]: (Dionisi, Prodolliet, & Tagliaferri, 1995); [10]: (Ourrach,
Rada, Prez-Camino, Benaissa, & Guinda, 2012); [11]: (Chabreyrie et al., 2008); [12]: (Von Baer et al., 2007); [13]: (Christoph, Rossmann, & Voerkelius, 2003); [14]: (Lees, 2003).
194
Table 3
Selected applications of food ngerprinting approaches for the matrices honey, olive oil and wine.
Matrix/objective of
investigation
Analytical
technique4
Honey
Botanical origin
H NMR
HPLC-UV
46(5)
Direct analysis
PCA, HCA, KNN, (Boffo, Tavares, Tobias, Ferreira, &

SIMCA, PLS-DA Ferreira, 2012)
20(5)
Extraction and clean-up

(phenolic compounds)
Shift correction,
normalization, 1st
derivative, mean-centering,
scaling
Integration (s/n b 3), scaling
to zero mean and unit
variance
1st and 2nd derivative,
SavitzkyGolay smoothing
bucketing, normalization
Homogenization
353(6)
118(4)
Extract with chloroform, pH

adjustment
Dilution with water, extraction
with chloroform
Dilution with distilled water
1075
Dilution with distilled water,

lyophilization, dilution with
different deuterated solvents
Direct analysis (HSSPME)
Normalization
PCA, GDA
Normalization
OPLS-DA, OPLSHCA, SIMCA

PCA, PLS-DA,
SVM
SIMCA, UNEQ,
POTFUN
FADA, PLS, PCA
GCMS
77(7)
Raman
374(2)
Direct analysis
MSC
NIR
111(2)
Direct analysis
1st, 2nd derivative, SNV
FT-IR
373(6)
Direct analysis
SNV, 1st, 2nd derivative
NIR
373(2)
H NMR
Homogenization, dilution with

distilled water
41(25) Dissolved in D2O
182(10)
FT-IR
150(5)
H NMR
Homogenization, dilution with

distilled water, centrifugation,
ltration
Dilution with distilled water
None, different derivatives,

SNV
Mean centering, unit
variance
None
SNV, 1st, 2nd derivative
Source
(Cavazza, Corradini, Musci &

Salvadeo, 2013)
PCA, MD-DA,
(Chen et al., 2012)
BP-ANN
PCA, O2PLS-DA, (Schievano et al., 2012)
SVM
PCA, PLS-DA
(Schievano et al., 2010)
PCA
71(5)
PCA, LDA
bucketing normalization,
DmodX
n.g.3
H NMR
Geographical origin
Multivariate
model6
250(5)
FT-IR
Olive oil
Botanical origin
(differentiation between
olive oil and other edible
oils)
Pre-processing6
H NMR
Adulteration
Sample preparation5
NIR
H NMR
Botanical, geo-graphical
origin
geographical origin
n1(m2)
(Etzold & Lichtenberg-Kraag,

2008)
(Lolli et al., 2008)
(Aliferis, Tarantilis, Harizanis, &

Alissandrakis, 2010)
(Pierna et al., 2011)
(Oliveri et al., 2011)
PLS
(Hennessy, Downey, & O'Donnell,

2010)
(Woodcock et al., 2009)
PCA, PLS-DA
(Consonni & Cagliani, 2008)
PLS-DA, GP,
PLS-GP
(Donarski et al., 2008)
PLS, FDA,
SIMCA
PLS-DA
FA,GDA
PCA, LS-SVM,
SVM, BP-ANN,
LDA, KNN
(Hennessy et al., 2008)
PCA, PLS-DA,
iPLS-DA, ECVA,
iECVA
PLS-DA
(Javidnia, Parish, Karimi, &

Hemmateenejad, 2013)
LDA
(Concha-Herrera, Lerma-Garcia,
Herrero-Martinez, & SimoAlfonso, 2009)
(Pizarro, Rodrguez-Tecedor,
Prez-del-Notario, Esteban-Dez,
& Gonzlez-Siz, 2013)
(Bevilacqua, Bucci, Magr, Magr,
& Marini, 2012)
Raman
1
H NMR
NIR
75150(2)
126(4)
135(2)
Direct analysis
Direct analysis
Homogenization
AirPLS
None
Smoothing, SNV
FT-IR
255(6)
Direct analysis
EMSC
PTR-MS
30(5)
Direct analysis
FT-IR
76(7)
Direct analysis
Autoscaling, meancentering
Normalization
UV/Vis
40(3)
Centrifugation
SNV, 1st derivative
PCA, LDA, PLSDA
NIR/MIR
57(4)
Direct analysis
PLS-DA, SIMCA
UV/Vis, NIR,
MIR
UV,NIR
57(2)
Centrifugation
120(3)
Direct analysis
H NMR
104(7)
Dissolving in CDCl3
Baseline correction, MSC,

1st, 2nd derivative,
combination of each
1st derivative,
normalization, SNV
Light scattering,
background noise, baseline
shift
Bucketing
SESI-MS
220(2)
Direct analysis
Normalization, meancentering, weighting
NIR
913(6)
Direct analysis
1st, 2nd Derivative, SNV
FT-IR
29(4)
Direct analysis
Derivative elaboration
GCMS
40(3)
Direct analysis (HS-SPME)
Synchronization retention
times, target peak
alignment
(Li et al., 2012)

(Bertelli et al., 2010)
(Zhu et al., 2010)
(Ruiz-Sambls et al., 2012)
PCA, UNEQ,
SIMCA, PLS
PLS-DA
(Casale et al., 2012)
PCA/CA/NCM
with MANOVA
PCA, LDA, PLSDA, KNN, CPANN
SIMCA, UNEQ,
POTFUN
CA, PCA, PLSDA
PCA, SLDA
(Longobardi et al., 2012)
(Lin, Chen, & He, 2012)
(Martnez-Lozano Sinues et al.,

2012)
(Oliveri et al., 2011)
(De Luca et al., 2011)
(Pizarro, Rodriguez-Tecedor,
Perez-del-Notario, & GonzalezSaiz, 2011)
195
Table 3 (continued)
Analytical
technique4
n1(m2)
Sample preparation5
Pre-processing6
Multivariate
model6
Source
122(6)
Drying, ltering, preparing

unsaponiable fraction
963(2)
Dissolving in CDCl3
PCA, LDA,
PLS-DA, CART,
SIMCA
PCA, LDA,
PLS-DA
(Alonso-Salces, Heberger, et al.,

2010)
(Alonso-Salces, Moreno-Rojas,
et al., 2010)
896(2)
Dissolving in DMSO, CDCl3
PLS-DA, SIMCA
(Mannina et al., 2010)
FT-IR
913(2)
Direct analysis
PCA, FDA, PLS
(Hennessy et al., 2009)
NIR
195(2)
Direct analysis
NIR
913(2)
Direct analysis
POTFUN,
SIMCA, UNEQQDA, MRM
PCA, PLS-DA
(Casale, Casolino, Ferrari, &

Forina, 2008)
Geographical origin, vintage

Geographical origin,
adulteration
Fluorescence 28 + 10 + 13 Centering
Dilution in
n-hexane
Phase- and baselinecorrection, normalization,

bucketing
Phase- and baselinecorrection, normalization,
bucketing
Baseline-correction,
normalization
1st, 2nd Derivative, meancentering, SNV
1st, 2nd Derivative,
detrending, SNV,
autoscaling, centering
None, 1st, 2nd derivative,
SNV
PCA, PLS
Adulteration
FT-IR
111(12) or
76(16)
Direct analysis
FT-IR
104(2)
Direct analysis
FT-IR
n.g.3
Direct analysis
NIR,Vis
136
n.g.3
FT-IR
44
Direct analysis
FT-IR
58
Direct analysis
FT-IR
58
Direct analysis
FT-IR
1
H NMR
59
92
Direct analysis
Dissolving in CDCl3
FT-IR
60
Direct analysis
DARTTOFMS
Raman
40
Dilution in toluene
Savitzky-Golay, 2nd
derivative
mean-centering
mean-centering, SNV
mean-centering, SNV
None, 1st, 2nd derivative
Phase- and baselinecorrection
Normalization, meancentering
Mass drift correction
35 + 48
Direct analysis
Normalization
PCA
ESI-FT-MS
72(6)
Dilution with methanol
Normalization
PCA, LDA
HPLC-TOFMS
GCMS
FT-MIR
1
H NMR
45(3)
Direct analysis
PCA, PLS-DA
75(3)
496
18(4)
PCA, SLDA
PCA, LDA
PCA, OPLS
(Zhang, Li, et al., 2010)

(Louw et al., 2009)
(Son, Hwang, Ahn, et al., 2009)
MIR
191(7)
Direct analysis (HSSPME)

Filtering
Drying, lyophilization, addition of
D2O and oxalate buffer
Direct analysis
Mean-centering and logtransformation

Autoscaling
Centering, scaling
Bucketing, normalization,
COW, mean-centering
None
(Villagra, Santos, Vaz, Eberlin, &

Felipe Laurie, 2012)
(Vaclavik et al., 2011)
PCA, LDA
Botanical origin, vintage,

quality
UPLC-FTICR-MS
400(34)
Direct analysis
HCA, PCA, LDA
(Bevin, Dambergs, Fergusson, &

Cozzolino, 2008)
(Cuadros-Inostroza et al., 2010)
Botanical, geographical origin
67(24)
Extraction of phenolic fraction
PCA, PLS-DA
(Anastasiadi et al., 2009)
Botanical, geographical
origin, authenticity
Geographical origin
574(57)
pH Adjustment
(Godelmann et al., 2013)
UV, IR
41(5)
Direct analysis
UV
35(3)
Addition of buffer
16(3)
pH Adjustment, centrifugation
PCA, LDA,
MANOVA, NCM
PCA, LDA,
SIMCA
PCA, HCA,
PLS-DA
PCA
110(3)
pH Adjustment, dilution with

water, lyophilization
FT-NIR
54(3)
Direct analysis
Matrix/objective of
investigation
H NMR
H NMR
H NMR
Authenticity, adulteration
Wine
Botanical origin
H NMR
H NMR
H NMR
H NMR
1st Derivative, baseline

correction, SavitzkyGolay
smoothing
Mean-centering and
orthogonally signal
corrected data
None, 1st, 2nd derivative
Normalization, logtransformation, scaling,

centering
Baseline correction,
bucketing, centering,
scaling
Bucketing
SNV, Smoothing, 2nd
derivative
Autoscaling
bucketing
normalization, meancentering, scaling
SNV, MSC, 2nd Derivative
(Kunz,
Ottaway,
Kalivas,
Georgiou, &
Mousdis, 2011)
HCA, PCA, PLSDA, PLS
(Woodcock et al., 2008)
(de la Mata et al., 2012)
PLS, PLS-DA,
VIP
(Oussama, Elabadi, Platikanov,

Kzaiber, & Tauler, 2012)
PCA, PLSR, PCR,

DA
PCA, PLS
(Rohman & Man, 2012b)

(Mignani et al., 2011)
PLS, PCR
(Rohman & Man, 2011)
PLS, PCR
(Rohman & Che Man, 2011)
PLS, PCR
(Rohman & Man, 2012b)
PLS, PCR, DA
(Rohman & Man, 2010)
PCA, LDA, MRM (Mannina et al., 2009)
PCA, PLS-DA
LDA
PCA, PLS-DA
PCA, DA,
PLS-DA
(Obaidat, Kahnfar, & Obaidat,

2009)
(Vaclavik, Cajka, Hrbek, &
Hajslova, 2009)
(Zou et al., 2009)
(Martelo-Vidal, Domnguez-Agis,
& Vzquez, 2013)
(Azcarate, ngel, Pellerano,
Marchevsky, & Camia, 2013)
(Koda, Furihata, Wei, Miyakawa,
& Tanokura, 2012)
(Papotti et al., 2013)
(Shen et al., 2012)

(continued on next page)
196
Table 3 (continued)
Analytical
technique4
n1(m2)
Sample preparation5
Pre-processing6
Multivariate
model6
Source
UV, Vis, NIR,

MIR
UV/Vis, NIR,
MIR
1
H NMR
64(2)
Direct analysis
n.g.3
98(5)
Direct analysis
(Cozzolino, Cynkar, Shah, &

Smith, 2011)
(Riovanto et al., 2011)
16(3)
Stirring
H NMR
20(4)
NIR, Vis
50(4)
Drying, lyophilization, addition of

D2O and oxalate buffer
Direct analysis
Geographical origin, vintage
111(37)
pH Adjustment, centrifugation
Adulteration
Authenticity
H NMR
MIR
n.g.
59(2)
ph Adjustment
Direct analysis
SNV, 2nd Derivative,

smoothing
centering
COW, mean-centering
SNV, 2nd Derivative
transformation, smoothing
Normalization, correction
of chemical shift, scaling
Bucketing
SNV
PCA, PLS-DA,
SIMCA
PCA, LDA,
SIMCA
PCA, DA, HCA
40(4)
FT-IR
65(3)
Extraction with ethyl acetate and

drying with Na2SO4
Concentration (lyophilization and
nitrogen-ow)
FT-IR
120
Direct analysis
NIR, Vis
20
Direct analysis
Matrix/objective of
investigation
H NMR
H NMR
PCA, PLS-DA
PCA, PLS-DA,
SLDA
PCA, ECVA,
iECVA
LDA, ANN
PCA, PLS
Mean-centering, scaling
PCA, PLS, OPLS
Transformation to
absorbance unit
PCA, CA, LDA,

CART
RegularizedDA
PCA
Normalization, 2nd
derivative
SNV, 2nd Derivative
transformation, smoothing
PLS
(Mazzei, Francesca, Moschetti, &

Piccolo, 2010)
(Son, Hwang, Kim, et al., 2009)
(Liu et al., 2008)
(Lopez-Rituerto et al., 2012)
(Imparato et al., 2011)
(Fudge, Wilkinson, Ristic, &
Cozzolino, 2013)
(Ali, Maltese, Toepfer, Choi, &
Verpoorte, 2011)
(Ioannou-Papayianni, Kokkinofta,
& Theocharis, 2011)
(Zhang, Chen, et al., 2010)
(Smyth et al., 2008)
1 n: number of samples; 2 m: number of groups; 3 n.g.: information not given.

4 Abbreviations: DARTTOF-MS: direct analysis in real time time-of-ight mass spectrometry; ESI-FT-MS: electrospray ionization Fourier transform mass spectrometry; FT-IR: Fourier
transform infrared spectroscopy; FT-NIR: Fourier transform near infrared spectroscopy; GCMS: gas chromatography hyphenated with mass spectrometry; 1H NMR: proton nuclear
magnetic resonance spectroscopy; HPLCTOF-MS: high-performance liquid chromatography coupled to time-of-ight mass spectrometry; HPLC-UV: high-performance liquid
chromatography in combination with an ultraviolet detector; IR: infrared spectroscopy; MIR: mid infrared spectroscopy; NIR: near infrared spectroscopy; PTR-MS: proton transfer reaction
mass spectrometry; SESI-MS: secondary electrospray ionization mass spectrometry; UPLC-FT-ICR-MS: ultra performance liquid chromatography hyphenated with Fourier transform ion
cyclotron resonance mass spectrometry; UV: ultraviolet spectroscopy; Vis: visible spectroscopy.
5 Abbreviations: CDCl3: deuterated chloroform; DMSO: dimethyl sulfoxide; D2O: deuterium oxide; HS-SPME: headspace solid phase micro extraction; Na2SO4: sodium sulfate.
6 Abbreviations: airPLS: adaptive iteratively reweighted penalized least squares; ANN: articial neural network; BP-ANN: back propagation articial neural network; CA: canonical
analysis; CART: classication and regression trees; COW: correlation optimized warping; CP-ANN: counter-propagation articial neural networks; DA: discriminant analysis; DmodX:
distance to the model in x-space; ECVA: extended canonical variable analysis; EMSC: extended multiplicative scatter correction; FA: factor analysis; FDA: factorial discriminant analysis;
GDA: general discriminant analysis; GP: genetic programming; HCA: hierarchical cluster analysis; iECVA: interval extended canonical variable analysis; iPLS-DA: interval-partial least
square discriminant analysis; KNN: k-nearest neighbor classication; LDA: linear discriminant analysis; LS-SVM: least square support vector machines regression method; MANOVA:
multivariate analysis of variance; MCCV: Monte-Carlo embedded cross-validation; MD-DA: Mahalanobis-distance discriminant analysis; MRM: multiple regressions models; MSC:
multiplicative scatter correction; NCM: nearest class mean; O2PLS-DA: orthogonal projection to latent structures discriminant analysis; O2PLS-HCA: orthogonal projection to latent
structures hierarchical cluster analysis; OPLS: orthogonal partial least squares; PCA: principal component analysis; PCR: principle component regression; PLS-DA: partial least squaresdiscriminant analysis; PLS-GP: projection to latent structures genetic programming; POTFUN: potential function techniques; SIMCA: soft independent modeling of class analogy; SLDA:
stepwise linear discriminant analysis; SNV: standard normal variate transformation; SVM: supported vector machine; UNEQ: unequal class spaces; UNEQ-QDA: unequal-quadratic
discriminant analysis; VIP: variable importance of projection.
2010). A detailed consideration of the respective mathematical models

is not in the perspective of this review and is already covered in
literature (Bevilacqua et al., 2013; Li Vigni, Durante, & Cocchi, 2013;
Trygg, Holmes, & Lundstedt, 2006; Varmuza & Filzmoser, 2009;
Westad, Bevilacqua, & Marini, 2013). All of these methods are applicable
for two-dimensional data sets. Only a few studies deal with the
multivariate analysis of three-dimensional data sets. Thereby, mainly
software packages developed by manufacturers of analytical equipment
are used (e.g. MarkerLynxTM software from Waters (Milford/USA),
MarkerviewTM software from AB SCIEX (Framingham/USA) or SIEVE
from ThermoFisher Scientic (Waltham/USA)) (Yoshida, Yamazaki,
Ozawa, Mizukoshi, & Miyano, 2009). Another solution for the analysis
of three-dimensional data is the parallel factor analysis (PARAFAC),
which was described by e.g. Matos et al., Pierce et al. and other standard
references as a chemometric data processing technique for comprehensive multi-dimensional chromatography data (Amigo & Marini, 2013;
Matos, Duarte, & Duarte, 2012; Pierce, Kehimkar, Marney, Hoggard, &
Synovec, 2012; Smilde, Bro, & Geladi, 2004). Furthermore, a PARAFAC
model was applied by Cocchi et al. to characterize and discriminate
different balsamic vinegar varieties based on headspace mass spectrometry (HS-MS) ngerprints (Cocchi, Durante, Marchetti, Armanino,
& Casale, 2007).
In addition to multivariate statistics, some studies use univariate data
analysis. With the background of the food scandal in 2008, Lachenmeier
et al. examined infant formula to detect melamine adulterations using

H NMR (Lachenmeier et al., 2009). Thereby, the statistical analysis was
based on the use of quantiles obtained from the distribution of reference
samples estimating the median and the corresponding variances at every
chemical shift. Melamine adulterations were then detected based on the
resulting z-scores (Lachenmeier et al., 2009). In this case, univariate data
analysis was successful to reveal the adulterated samples. In general,
however, multivariate data analysis of ngerprints has some advantages
compared to that of univariate statistics of e.g. single components of
the ngerprints (Olivieri, 2008; Trygg, Gullberg, Johansson, Jonsson, &
Moritz, 2006), which might be the rst instance that the whole spectra
of information detected with the non-targeted approach is used.
Multivariate data analysis often starts by applying unsupervised
techniques such as PCA, for data exploration. Thereby, PCA focuses on
the maximum variation in the data set of different food samples.
These largest variations, however, are not necessarily caused by
variations in the food composition only, but may also occur due to
variances during sample preparation or analytical measurement,
e.g. by using analytical equipment from different suppliers. If such sources
of variances play a relevant role in the overall variation of the data set
and are not known, the interpretation of the multivariate data analysis
might be confounding and even lead to a false differentiation of the
respective food samples. Therefore, in non-targeted ngerprinting,
quality control measures are essential to follow and account for possible
1
197
procedure
Data
evaluation
Analytical
procedure
tru
m
en
t
va al
ria
tio
n
Standardized
in
s
pre-processing
te
m
Fingerprint
collection
po
ra
l
va
ria
tio
n
Consistent
Sample
preparation
Standard
measurement
procedure
Fig. 2. Requirements for food ngerprinting approaches for the application in food control.
variances caused by both, sample preparation and analytical measurements. The issue of quality assurance for applying food ngerprinting
in food control is a central topic in this review continued in chapter 2.3.
2.2. Fingerprinting techniques for food authentication status quo

The number of research studies on food ngerprinting published in
peer-reviewed journals so far is about 350. Since this article gives an
overview on the most common applications of food ngerprinting,
three matrices were selected, which were most often investigated:
honey, olive oil and wine. For a comprehensive overview, Table 2
describes the typical objectives in food authentication (geographical
or botanical origin and adulteration) of these three matrices as well as
exemplarily targeted approaches tackling some of the aims. To investigate
the geographical origin, e.g. stable isotope ratio mass spectrometry
(hydrogen, carbon, oxygen and nitrogen) or the quantitative analysis of
trace elements (e.g., Al, Ca, Cr, Fe, K, Li, Mn and Na) by e.g. inductively
coupled plasma atomic emission spectroscopy (ICP-AES) or atomic
absorption spectroscopy (AAS) are typically used (Drivelos & Georgiou,
2012). The analytical parameters for the determination of the botanical
origin as well as potential adulteration are in general matrix specic: in
case of honey, the microscopic investigation of pollen is often used,
whereas triglyceride or fatty acid proles are investigated in olive oil.
Furthermore, it becomes obvious that the individual objectives
could rarely be studied with a single measurement or parameter by
using classical targeted approaches. This, however, is expected for the
non-targeted ngerprinting analysis and was already successfully
demonstrated, for instance, for the combined investigation of the
variety and geographical origin of apple juices by chromatographic
ngerprints (Guo et al., 2012). Further, by an investigation of the
shikimic acid content only the identity of Burgundy wines can be
veried with certainty. For other white wines, there is currently no
method available for the differentiation between varieties in forensic
applications. Hence, the potential of non-targeted food ngerprinting
for closing methodological gaps in authenticity testing and for
investigating multiple objectives with only one analytical method are

clearly advantageous over the classical targeted approaches.
All three matrices are in a liquid state, more or less, so that they are
easily accessible to many analytical techniques without (or with only
little) sample preparation. Furthermore, these matrices play an important
role in food control in terms of authenticity and possible adulteration
issues (Moore, Spink, & Lipp, 2012). Table 3 gives a comprehensive
overview on food ngerprinting research studies using these matrices.
2.2.1. Honey
Honey is an extensively investigated food matrix. Primarily, honey
consists of an aqueous solution (with a water content of about
20%) of invert sugar. The predominant sugars are fructose and glucose
in approximately equal amounts (about 3138%). Furthermore, honey
consists of small quantities of organic acids, amino acids, minerals,
enzymes and pollen grains derived from honey collection. As a 100%
natural product, honey is produced by bees (Apis mellifera) and has to
be free from organic or inorganic matters foreign to its natural composition (Codex Alimentarius, 2001). As honey is a relative expensive
commodity, it is susceptible to fraud, e.g. by the extension with cheaper
sources of sugar. Therefore, Bertelli et al. developed a method using 1H
NMR to detect adulteration in honey falsied by the intentional addition of different concentrations of commercial sugar syrups (Bertelli
et al., 2010). The revealed data sets with authentic and adulterated
honey spectra were statistically analyzed by factor analysis (FA)
and GDA, whereas FA was used to reduce the raw data set. As a
result, the GDA model was judged to be able to identify sufciently
separated clusters corresponding to the authentic and adulterated
honeys; furthermore, it was possible to discriminate between different
adulteration levels. Also Li et al. as well as Zhu et al. investigated
successfully the non-targeted detection of adulteration (syrups and
sweeteners) in honey using vibrational spectroscopy (Raman and
NIR) (Li, Shan, Zhu, Zhang, & Ling, 2012; Zhu et al., 2010).
A further priority area of food ngerprinting investigations was set
on the botanical and geographical origin of honey. The differentiation
of monooral Italian honey types (e.g. acacia, orange, eucalyptus and
198
chestnut) and subsequent identication of specic key components

for each monooral origin were a particular aspect of these studies
(Lolli et al., 2008; Schievano, Peggion, & Mammi, 2010; Schievano
et al., 2012). The statistical data analysis was performed using both
unsupervised (PCA) and supervised models (GDA, partial least squaresdiscriminant analysis (PLS-DA) and orthogonal projection to latent
structures discriminant analysis (O2PLS-DA)), whereby the supervised
techniques demonstrated a high efciency for the prediction of botanical
origin. The investigations of Lolli et al. resulted in unacceptable results of
group samples as a function of their botanical origin using 1H NMR
(Lolli et al., 2008), whereas Schievano et al. obtained a reasonable
discrimination (Schievano et al., 2010). This might be due to different
sets of samples (with distinct intrinsic variations) and/or different preprocessing techniques as well as the potentially limited meaning of PCA
in this case.
A further important aspect in food authenticity, including the matrix
honey, as well as olive oil and wine, is the consideration of Protected
Designation of Origin (PDO) products. This declaration describes the
geographical origin of the product, strengthens the bond with the
territory of origin and promotes the consumers' perception of greater
protection and quality.
The analysis for monitoring specic geographical origins is challenging for analysts in principle, because there is usually no single parameter (chemical or physical) available, which allows a comprehensive
statement. Therefore, the development of global analysis approaches,
such as food ngerprinting, are especially promising for determining
geographical authenticity. As already mentioned, the EU project TRACE
intended to investigate olive oil as well as honey in concern to their
geographical origin using non-targeted vibrational spectroscopy. In addition to respective studies mentioned in the context of the research project
TRACE, for example, Consonni and Cagliani distinguished Hungarian and
Italian acacia honey by 1H NMR ngerprints (Consonni & Cagliani, 2008).
Here, the instrumental analysis was followed by a suitable pre-processing
(e.g. mean centering and unit variance scaling) and subsequent PCA as
well as PLS-DA analyses.
2.2.2. Olive oil
Similar to honey, virgin and extra virgin olive oils are high price
commodities, because of their production process (only by mechanical
means) combined with corresponding nutritional properties and
sensorial quality. As a consequence, this led to adulterations of
(extra) virgin olive oils with low-grade foreign oils of different
botanical origin (e.g. seed oil) or rened olive oils in the past
(Ogrinc, Kosir, & Spangenberg, 2003). Furthermore, there is also a
need of analytical techniques, which are able to discriminate olive oils
of different geographical origins as some of these products are assigned
to protected geographical indication labels (European Community,
1992; Le Gall & Colquhoun, 2003). Consequently, most of the studies
shown in Table 3 investigated possible strategies to characterize extra
virgin olive oil according to their geographical origin. In particular, the
studies of Alsonso-Salces et al. and Mannina et al. should be highlighted
(Alonso-Salces, Moreno-Rojas, et al., 2010; Mannina, Marini, Gobbino,
Sobolev, & Capitani, 2010). These studies showed excellent results in
correct sample classication/prediction (up to 90%) using 1H NMR and
subsequent PLS-DA. Due to its immense number of samples (N 895),
investigations of Alsonso-Salces et al. and Mannina et al. provided the
basis for developing a robust database. As already mentioned above
(Section 2.1 general approach) olive oil was also the subject of
investigations within the EU-funded project TRACE. FT-IR as well as
NIR spectroscopy were performed directly (without any sample preparation) to discriminate 913 samples between Ligurian and nonLigurian olive oils (Hennessy et al., 2009; Oliveri et al., 2011; Woodcock
et al., 2008). The investigated samples from three harvest periods
(20052007) were derived from Italy, France, Spain, Greece, Cyprus and
Turkey, whereby approximately one-fourth of all samples originated
from Liguria. Oliveri et al. compared different class-modeling techniques
to identify an adequate strategy to evaluate this unequally distributed

sample set (Oliveri et al., 2011). Here, SIMCA, UNEQ as well as POTFUN
were used to predict the object class of a test data set with so-called
one-class models (developed using objects of a training data set). In this
study, POTFUN was the best mathematical model for Ligurian oils
(efciency around 83%, whereas efciency has been computed in this
study as geometric mean of sensitivity and specicity values). In
comparison, Woodcock et al. used the same sample sets and the
same analytical technique (NIR) performing PLS-DA, whereas a dummy
Y-variable was assigned (1 for Ligurian and 0 for non-Ligurian olive oil
samples) resulting in a two-class model (Woodcock et al., 2008). The
correct classication rates achieved by PLS-DA (92.8% and 81.2% of
Ligurian and non-Ligurian samples in test set, respectively) are comparable to those calculated by POTFUN. Hence, based on the studied sample
sets and the applied analytical procedure, the assignment success for
Ligurian and non-Ligurian olive oils seems independent of the type of
pre-processing and the mathematical model used (one or two class
model). The differentiation between geographical origins was also
studied by Longobardi et al. using 1H NMR ngerprints (Longobardi
et al., 2012). Here, after data reduction by bucketing and PCA, the class
separation was rened by canonical analysis, then applied using the
nearest class mean and validated by the Monte-Carlo embedded crossvalidation, which together revealed a satisfying prediction of the origin
of olive oil samples.
The identication of intentional falsication of high-priced coldpressed olive oil with edible oils of neutral taste or olive oils of lower
quality is a common objective in the authenticity control of edible oils.
This is also reected in the number of applications of food ngerprinting
approaches studying this issue (Table 3). Abdul Rohman and colleagues
investigated in several studies the applicability of food ngerprinting for
the identication of illegal additives in olive oil by FT-IR. The issues of
these studies were inter alia identifying the quantities of to olive oil
added palm oil (Rohman, Che Man, Ismail, & Hashim, 2010; Rohman
& Man, 2010), lard (Rohman, Che Man, Hashim, & Ismail, 2011), rice
bran oil (Rohman & Man, 2012b), corn oil and sunower oil (Rohman
& Che Man, 2012). Here, samples were analyzed directly without
sample preparation and various pre-processing techniques were tested
(including rst and second derivatives, mean-centering, SNV). The
statistical analysis of the obtained data was performed using PCA, PLS
and principal component regression (PCR) to quantify successfully the
level of adulteration. Noteworthy in this context is another feasibility
study, which focuses on the differentiation between olive oils of different quality categories as well as varieties and other edible oils, such
as sunower, corn, soybean, canola oils and respective blends (de la
Mata et al., 2012). Here, de la Mata et al. also used FT-IR without any
sample preparation. Several unsupervised (HCA and PCA) and supervised
(PLS-DA) classication as well as regression (PLS) models resulted in
evident characteristics: 100% correct classication was calculated for
olive oil samples within a collection of pure samples of different types
of vegetable oils including olive oil of different categories, varieties and
origin. Furthermore, it was possible to differ between blends with olive
oil content higher than 50% (w/w) and below 50% (w/w). Another
study within the European MEDEO project (Methods and Tools for
dual access to the EO databases of the EU and Russia) from Mannina
et al. investigated the detection of rened hazelnut oils in admixtures
with rened olive oils using 92 oil samples of different origins by 1H
NMR (Mannina et al., 2009). Among other statistical techniques, the
analysis was performed by multiple regressions models (MRM), showing
regressions of R2 N 0.9984, which indicates a potential identication of
hazelnut oil contamination in olive oil up to 10%.
2.2.3. Wine
The matrix wine is of particular importance within the authenticity
testing of foods, because it has always been subject to various distortions and deceptions. Wine production and trade has always been
associated with high costs, which made it vulnerable for adulteration
to reduce these costs and achieve higher prots. For instance, the nonauthorized addition of sugar or concentrated grape must before or
during fermentation is misused to increase the natural ethanol content
and thus the value of the wine, for which a higher market price might be
achieved (Ogrinc et al., 2003). Another example is the addition of lowquality wines to wines of higher quality to obtain higher prot. For
this reason, regular and thorough state-of-the-art controls of wine are
needed to ensure the quality and authenticity of the products using
appropriate analytical procedures. As shown in Table 3, wine was
investigated using spectroscopic and/or spectrometric food ngerprinting techniques to identify their respective botanical and/or geographical
origin as well as possible adulterations. Imparato et al. developed an
analytical procedure using 1H NMR to enable the identication of red
wine adulterated with glucose syrup (Imparato et al., 2011). The nontargeted NMR analysis of aqueous matrices, such as wine, is especially
challenging due to the dominant water signal as well as ethanol signals
in the ngerprints masking interesting information. To expand the
dynamic range for the analysis of minor elements, a common method is
to suppress the water/ethanol signals using specic pulse sequences as
shown in the study from Monakhova et al. (Monakhova et al., 2011).
Alternatively, water and ethanol contents were reduced by lyophilization
and subsequent dissolution in deuterated solvents, e.g. deuterium
oxide (Papotti et al., 2013; Son, Hwang, Ahn, et al., 2009), which,
however, goes along with an intense sample preparation as well
as an unnatural modication of the matrix. The rst applications on
the discrimination between wine samples derived from different
grape varieties (Anastasiadi et al., 2009; Son, Hwang, Ahn, et al., 2009)
were largely based on 1H NMR and PCA. The identication of key
compounds, which contributed to the mathematical discrimination of
wines, was typically achieved by the inspection of the corresponding
loadings. Although the two studies showed different key compounds,
probably due to divers sample sets, both could demonstrate the high
application potential and feasibility of food ngerprinting using 1H
NMR.
Zhang et al. investigated discrimination between varieties of Chinese
wine by HS-SPME coupled to a gas chromatographic separation of
matrix and subsequent mass spectrometric detection (Zhang, Li,
et al., 2010). For this purpose, samples of the varieties Cabernet
Sauvignon, Merlot and Cabernet Gernischt were directly analyzed to
determine volatile compounds. Thirty major peaks were chosen for data
evaluation, starting with analysis of variance (ANOVA) to exclude nonsignicant signals. PCA and stepwise linear discriminant analysis (SLDA)
were used as multivariate models to identify volatile compounds (mainly
ethyl esters), which enabled the differentiation and classication of
samples.
A further study on wine aimed to distinguish several wine attributes
(variety, vineyard, vintage and quality) within about 400 commercial
wine samples from Chile. The investigations based on a non-targeted
approach using ultra performance liquid chromatography (UPLC)
hyphenated to a FT-ICR-MS (Cuadros-Inostroza et al., 2010). The rst
data exploration suggested that most of the variation in the data set
might be explained by the grape varieties. As a consequence, the sample
set required a higher homogeneity to discriminate additional features
like vineyard or quality. Therefore, the authors used a kind of hierarchical
order (according to the decreasing variance explained in the data set) to
investigate the different objectives.
PDO products play an important role in the wine sector. Recently,
Papotti et al. successfully used 1H NMR spectroscopy to characterize and
assess the authenticity of PDO Lambrusco wine, by differentiating several
categories: Lambrusco di Sorbara, Lambrusco Salamino di Santa Core and
Lambrusco Grasparossa di Castelvetro (Papotti et al., 2013). Additionally,
a similar aim was investigated by Riovanto et al. (determination of the
geographical origin of Australian Shiraz wines from different regions)
using multiple spectroscopic methods UVvis, NIR and MIR (Riovanto,
Cynkar, Berzaghi, & Cozzolino, 2011). Here, it was observed that the
best classication results were obtained using MIR followed by a
199
statistical analysis of data using SIMCA (overall correct classication of

95.3%).
2.3. Challenges in ofcial food control
The classical assessment of authenticity by means of indicator
analytes and their comparison with reference data requires, apart
from the analytical expertise, specic comprehensive product knowledge
needed for the interpretation and objection. In addition, it is of utmost
importance that analytical results are comparable to those obtained in
different laboratories. Therefore, applicable and validated standard
methods play an important role. Only the high requirements to the
analytical accuracy, respective trueness and precision, enable the
establishment of reference data and the comparison with other
laboratories. Exchangeability and comparability of data is one of the
essential key points in ofcial food control. Up to now, this is only
realized by dened analytical parameters, which stands in contradiction
to the multivariate spectroscopic or spectrometric data generated by
ngerprinting techniques.
2.3.1. Data consistency
Most research studies employing the multivariate analysis of
food ngerprints are so-called feasibility studies, which investigate
differences or common properties within one set of samples of a
given commodity. Further investigations of the same sample matrix
under standardized conditions with e.g. new or extended sample pools
or varying research questions are rare, so that the general conclusions
for certain commodities on the relevance of inuencing factors are
limited. Moreover, these feasibility studies are usually undertaken within
one laboratory on one instrument exclusively, which restricts validation
possibilities of statistically-derived results of the non-targeted procedure.
the validation of the whole analytical procedure including statistical
data evaluation and consistency of the measurement over time,
instruments and laboratories is, however, essential in ofcial food
control (Fig. 2). The application of these non-targeted ngerprinting
approaches is due to actual missing validation strategies still restricted
for ofcial control purposes, thereby offering complex challenges for
the scientic community.
However, two practical ngerprinting applications in the eld of
food authentication should be mentioned here as examples overcoming
most of these limitations by implementing a standardized operating
protocol. The joint efforts of Bruker BioSpin GmbH and SGF International e.V. resulted in NMR-based screening methods, so-called FoodScreener for fruit juices (JuiceScreener) and wine (WineScreener)
combined with Proling technique (Bruker BioSpin, 2012; Godelmann
et al., 2013; SGF International e.V., 2012; Spraul et al., 2009). Herein, it
is possible to compare non-targeted spectral 1H NMR data with the
corresponding group of reference spectra (e.g. database of several
thousands of reference juices, obtained from production sites all over
the world) using verication models. The aim of the classication analysis
in the case of the JuiceScreener is the determination of the origin of
the fruit, its variety (botanical origin) and the differentiation between
juice made from concentrate and juice not from concentrate. These
applications unify measurements from different instruments of the
same type and vendor by using a standardized operating procedure
(including sample preparation and measurement protocol), which allows
evaluation of the data with the same statistical models. However, in these
cases the measurement procedures, the reference data (data base), its
assessment and evaluation, are marketed as intellectual properties. The
use or the acceptance of such proprietary methods or measurement
procedures in standard setting organizations (Codex Alimentarius as
well as Organization International Du Vin, OIV) is limited at this stage
and are therefore currently not used in ofcial consumer protection.
Ofcial food control implies the application of valid (analytical) test
procedures enabling different laboratories testing the same sample to
obtain the same analytical results. However, validation procedures
200
and respective classical parameters according to established procedures

e.g. 2002/657/EC Commission decision of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical
methods and the interpretation of results (Commission of the European
communities, 2002) are so far not suited for non-targeted approaches.
Therefore, new strategies and terminologies must be dened. In particular, transferability of the models and/or the data to other facilities
is typically not considered, when models are created on a certain
commodity within one laboratory with one particular instrument.
However, the transfer to other instruments of the same type and vendor
or, even more challenging, to other instruments from different vendors
is essential in ofcial food control. The requirements are especially
linked to reference data, which has to be traceable, readable, useable
and expandable (TRUE data (Guillou, 2012)) in a long-term perspective. This means that the data created need to have a certain unique
format, which remain constant for a long term. Initial suggestions on
data exchange standards were proposed by a working group of the
Metabolomics Standards Initiative (Fiehn et al., 2007; Hardy & Taylor,
2007). To realize data exchange in food control, input from different
sites would be needed here: product knowledge and its application
in control, the platform used, and the analytical technique chosen.
Furthermore, ideally different vendors should cooperate together and
work out sustainable solutions.
Hence, to enable the application of non-targeted ngerprinting
approaches in ofcial food control, transferability of the data is essential.
For this, standard data formats and operation procedures to be followed
strictly throughout the whole ngerprinting procedure need to be
established and moreover, validation standards comparable to classical
validation strategies need to be developed.
2.3.2. Quality assurance
The implementation of quality assurance measures for performance
control is fundamental in analytical chemistry and in particular essential
in ofcial food control in terms of the warranty and documentation of
analytical results. The claim of validity of ofcial investigation results
is raised as in case of an incident, the data also need to endure in
court. The aim is therefore to identify reliable analysis results, whose
accuracy is frequently tested and documented. For this, requirements
(quality attributes) need to be dened and quality assurance measures
need to be implemented in routine analysis.
In targeted analytical approaches, appropriate (internal and external)
measures are known and routinely used to control the requirements
previously determined during validation of a particular analytical
method. This includes internal laboratory activities, e.g. the regular
investigation of a quality control (QC) sample and the documentation
of its results in a control chart. A QC sample usually is a (certied or
in-house) reference material with a known content of the respective
compound or class of compounds. It is analyzed in parallel to the
routine measurement to check the performance of the whole analytical
procedure (from sample preparation to data evaluation). In addition to
that, external laboratory control is essential to guarantee long-term
management of the quality of analysis as well as the comparability of
results between different laboratories. A common approach is the
regular participation in inter-laboratory comparison studies (prociency
tests). Here, subsamples of a homogeneous sample are distributed by
the reference laboratory to participating laboratories, which analyze the
matrix with their own established analytical method to determine a
quantity value, then sending their results back to the organizer for nal
performance comparison across laboratories.
In the eld of non-targeted food analysis, such QC measures as well
as quality attributes have to be developed and derived to establish these
approaches as tools in food control. However, the implementation of QC
measures for ngerprinting approaches is complex, since not a single
quantitative data, but the full spectrum of the food matrix needs to be
considered. Current food ngerprinting applications typically do not
consider quality assurance measure comparable to classical analytical
standards in their studies yet. For example, the use of QC samples is

usually omitted so far. Among recent food ngerprinting publications
only very few studies reported on the use of QC samples for each
measurement day to monitor the quality of the analytical measurement
and the data analysis procedure (Godelmann et al., 2013; Nietner, Pster,
Glomb, & Fauhl-Hassek, 2013). The importance of quality assurances
for the non-targeted approaches has also been raised in other elds of
metabolomics applications providing detailed considerations on the
purpose and usage of QC samples depending on the size of the study
(Dunn, Wilson, Nicholls, & Broadhurst, 2012).
In food ngerprinting, two main challenges have to be addressed
to implement QC samples: the nature of the sample itself and the
statistical approach to derive adequate quality attributes. So, the QC
sample should fulll certain criteria: as in targeted analysis it should
be homogeneous so that aliquots of the sample are comparable
throughout its usage. Furthermore, it also should sufciently reect
the properties of the investigated matrix to enable quality control
from sample preparation to data analysis. For this, a similar matrix as
the study matrix seems advantageous. If not purchasable, pooled QC
samples as suggested for biouids (Dunn et al., 2012), might be prepared
by combining aliquots of e.g. all samples of a particular experiment to
cover all signals contained in the matrix. Moreover, the QC sample should
have a certain shelf life and thus be stable over the observation period.
Compared to targeted analysis, where only the analyte or a group of
compounds of interest has to be free of decomposition or degradation,
in food ngerprinting, the stability of the matrix itself has to be ensured.
This is difcult because most foods are subject to alteration and spoilage
over a certain period of time. For example, edible oils themselves might
oxidize even under freezing conditions or wine spoils due to microbial
and/or enzymatic inuence. As a consequence, such matrices might
only be used for a dened period and the QC sample might then be
renewed with a certain observation time overlap. Additionally, the
matrix-like QC sample might be accompanied by a chemically clearly
dened sample related to the matrix. Nietner et al., for instance, used
starch, in addition to their QC sample of investigated feed material,
which is a main component in the study matrix and certainly stable for
a longer time (Nietner et al., 2013). So, storage-related deviations in the
matrix-like QC sample might additionally be followed.
The second challenge of implementing a QC sample is the denition
of a decision criterion for the quality of non-targeted analysis. A control
chart in classical sense is not applicable here, since the complete spectrum
of the matrix should be considered. A reasonable visual inspection of the
matrix-like QC sample might be provided with the scores plot of the
PCA, including both, the study samples and QC samples as suggested by
Sangster et al. (Sangster, Major, Plumb, Wilson, & Wilson, 2006). With
that approach, the relevance of analysis-related variation to the differentiation problem of the study might be observed. The analytical procedures
might be considered adequate for the specic study, if the cluster of QC
samples is small compared to the clusters of within-group and
between-group food samples. This is useful for internal study
control, but does not allow comparison, e.g. with different approaches
as no study-independent criteria could be derived from that. Furthermore, the data evaluation is performed only visually, and no numerical
decision criteria could be specied. Therefore, an alternative approach is
to be chosen, including a dened range, in which the analytical results
of QC sample are allowed to vary.
A valuable measure for external laboratory quality assurance might
be inter-laboratory comparisons based on ngerprints of the same
sample set detected with non-targeted approaches. Here, two options
might be thought of, one which only includes the quality of the
analytical measurement and another which additionally comprises the
data evaluation. In the last case, the samples would have to be
statistically evaluated by the participating laboratories themselves,
only providing the nal e.g. classication or calibration result to the
organizer. This, however, would require the availability of an adequate
statistical model (including a sufcient number of training set samples)
in each laboratory. This has potentially to be provided by the organizer

together with the sample set. In the other case, the samples would be
only measured in each laboratory in a non-targeted manner under
conditions as comparable as possible and the raw data would have to
be sent to the organizer, who globally evaluates the measurement
performances. Such inter-laboratory comparisons are rare in the eld
of food ngerprinting and also in other application elds. An initial
study by Ward et al investigated the instrument comparability of
non-targeted ngerprints of plants derived with NMR devices from
the same provider across different sites and different magnetic eld
strengths (400, 500 and 600 MHz) (Ward et al., 2010). As example,
the ngerprints of two different genotypes of broccoli were collected
in each laboratory and processed with multivariate data analysis
within one PCA model. Interestingly, the two genotypes of broccoli
could be clearly distinguished irrespective of the instrument location,
but the genotype variation was less intense than the inuence of the
eld strength, thus the spectral resolutions, on the ngerprints, which
could, however, be adjusted by data preprocessing. This study exemplarily showed the robustness of a non-targeted approach across laboratories and further, the feasibility of laboratory comparison studies for
performance control of specic differentiation problem. Here again,
the statistical evaluation of comparability relies on the variation of the
overall data set. For more general evidence, however, clearly dened
decision criteria beyond visual inspection of PCA, independent from the
respective study designs would be essential and need to be developed
for adequately validating a non-targeted approach with perspective use
in ofcial food control.
3. Conclusion
The applications of the ngerprinting techniques respectively seem
to be a useful and powerful tool in view of comprehensive food control.
Many investigations have shown the applicability of these non-targeted
strategies for various foodstuffs. There is a range of analytical methods
available including spectroscopic techniques (e.g. NMR, MIR and NIR)
as well as mass spectrometry (partially coupled with chromatographic
separation). The subsequent statistical data analysis (optionally multivariate data analysis) allows an identication of many deviations from
the expected product. Besides the authenticity testing of food (botanical
or geographical origin, species, etc.) a comprehensive analysis based on
chemical ngerprints enables also the detection of hazardous or safetyrelated manipulations (or violations) within the meaning of the existing
food law in the form of unauthorized additives or the use of banned
technological processes. Due to this high potential in applications, this
approach will be expanded and improved further.
Apart from the numerous opportunities and benets of the food
ngerprinting approaches, these applications show also limitations,
which need to be overcome prior to the potential implementation of
these innovative approaches in food control. Particular challenges are
the development and establishment of standardized/harmonized operating procedures as well as quality assurance measures, for instance,
the standardization of chemical and statistical analysis, the validation
(including e.g. test of robustness and precision) of the entire process
and the introduction of reference standards. Standardized reporting in
food ngerprinting comparable to the minimal reporting standards
of other application elds under the umbrella of the Metabolomics
Standardization Initiative might be a rst step to enhance transparency
and interpretability of study results in a more global context. This needs
then to be supported by easily accessible data exchange formats with a
uniform output, which allow broad chairing of acquired raw data, but
moreover the development and establishment of joint comprehensive
databases. Further, the development of appropriate validation strategies
is essential to ensure valid and reliable analytical results. Thereby, not
only does an appropriate validation of the statistical models need to
become a natural routine, but also procedures that need to be thought
of to validate the overall analytical process. For this, quality assurance
201
measures might be adopted from the classical analytical chemistry such

as quality samples or prociency tests.
Hence, this review shows not only the potential of the non-targeted
food ngerprinting approach, but also the challenges of this complex
procedure for its potential application in ofcial food control. This
makes clear that joint effort is necessary in ofcial food control to
provide solutions for the particular challenges and to furthermore
establish these techniques for the consumer protection and/or for
quality control purposes.
References
Ali, K., Maltese, F., Toepfer, R., Choi, Y. H., & Verpoorte, R. (2011). Metabolic
characterization of Palatinate German white wines according to sensory attributes,
varieties, and vintages using NMR spectroscopy and multivariate data analyses. Journal
of Biomolecular NMR, 49, 255266.
Aliferis, K. A., Tarantilis, P. A., Harizanis, P. C., & Alissandrakis, E. (2010). Botanical
discrimination and classication of honey samples applying gas chromatography/mass
spectrometry ngerprinting of headspace volatile compounds. Food Chemistry, 121,
856862.
Alonso-Salces, R., Heberger, K., Holland, M. V., Moreno-Rojas, J., Mariani, C., Bellan, G.,
et al. (2010). Multivariate analysis of NMR ngerprint of the unsaponiable fraction
of virgin olive oils for authentication purposes. Food Chemistry, 118, 956965.
Alonso-Salces, R. M., Moreno-Rojas, J. M., Holland, M. V., Reniero, F., Guillou, C., &
Heberger, K. (2010). Virgin olive oil authentication by multivariate analyses of H-1
NMR ngerprints and delta C-13 and delta H-2 data. Journal of Agricultural and Food
Chemistry, 58, 55865596.
Amigo, J. M., & Marini, F. (2013). Chapter 7 Multiway methods. In M. Federico (Ed.),
Data handling in science and technology chemometrics in food chemistry, Vol. 28.
(pp. 265313): Elsevier.
Anastasiadi, M., Zira, A., Magiatis, P., Haroutounian, S. A., Skaltsounis, A. L., & Mikros, E.
(2009). H NMR-based metabonomics for the classication of Greek wines according
to variety, region, and vintage. comparison with HPLC data. Journal of Agricultural and
Food Chemistry, 57, 1106711074.
Anklam, E. (1998). A review of the analytical methods to determine the geographical and
botanical origin of honey. Food Chemistry, 63, 549562.
Antignac, J. P., Courant, F., Pinel, G., Bichon, E., Monteau, F., Elliott, C., et al. (2011). Mass
spectrometry-based metabolomics applied to the chemical safety of food. TrAC
Trends in Analytical Chemistry, 30, 292301.
Aursand, M., Standal, I. B., Prael, A., McEvoy, L., Irvine, J., & Axelson, D. E. (2009). C-13 NMR
pattern recognition techniques for the classication of Atlantic salmon (Salmo salar
L.) according to their wild, farmed, and geographical origin. Journal of Agricultural
and Food Chemistry, 57, 34443451.
Azcarate, S. M., ngel, Pellerano, R. G., Marchevsky, E. J., & Camia, J. M. (2013).
Classication of Argentinean Sauvignon Blanc wines by UV spectroscopy and
chemometric methods. Journal of Food Science, 78, C432C436.
Berg, R., Hoefsloot, H., Westerhuis, J., Smilde, A., & Werf, M. T. (2006). Centering, scaling,
and transformations: Improving the biological information content of metabolomics
data. BMC Genomics, 7, 115.
Bertelli, D., Lolli, M., Papotti, G., Bortolotti, L., Serra, G., & Plessi, M. (2010). Detection of
honey adulteration by sugar syrups using one-dimensional and two-dimensional
high-resolution nuclear magnetic resonance. Journal of Agricultural and Food Chemistry,
58, 84958501.
Bevilacqua, M., Bucci, R., Magr, A.D., Magr, A. L., & Marini, F. (2012). Tracing the origin of
extra virgin olive oils by infrared spectroscopy and chemometrics: A case study.
Analytica Chimica Acta, 717, 3951.
Bevilacqua, M., Bucci, R., Magr, A.D., Magr, A. L., Nescatelli, R., & Marini, F. (2013).
Chapter 5 Classication and class-modelling. In M. Federico (Ed.), Data handling
in science and technology chemometrics in food chemistry, Vol. 28. (pp. 171233):
Elsevier.
Bevin, C. J., Dambergs, R. G., Fergusson, A. J., & Cozzolino, D. (2008). Varietal discrimination of Australian wines by means of mid-infrared spectroscopy and multivariate analysis. Analytica Chimica Acta, 621, 1923.
Boffo, E. F., Tavares, L. A., Tobias, A.C. T., Ferreira, M. M. C., & Ferreira, A. G. (2012).
Identication of components of Brazilian honey by 1H NMR and classication of its
botanical origin by chemometric methods. LWT Food Science and Technology, 49,
5563.
Bruker BioSpin (2012). JuiceScreener Enabling SGF Proling. http://www.brukerbiospin.com/juicescreener-dir.html
Caetano, S., stn, B., Hennessy, S., Smeyers-Verbeke, J., Melssen, W., Downey, G., et al.
(2007). Geographical classication of olive oils by the application of CART and SVM
to their FT-IR. Journal of Chemometrics, 21, 324334.
Casale, M., Casolino, C., Ferrari, G., & Forina, M. (2008). Near infrared spectroscopy and
class modelling techniques for the geographical authentication of Ligurian extra
virgin olive oil. Journal of Near Infrared Spectroscopy, 16, 16.
Casale, M., Oliveri, P., Casolino, C., Sinelli, N., Zunin, P., Armanino, C., et al. (2012).
Characterisation of PDO olive oil Chianti Classico by non-selective (UVvisible, NIR
and MIR spectroscopy) and selective (fatty acid composition) analytical techniques.
Analytica Chimica Acta, 712, 5663.
Cavazza, A., Corradini, C., Musci, M., & Salvadeo, P. (2013). High-performance liquid
chromatographic phenolic compound ngerprint for authenticity assessment of
honey. Journal of the Science of Food and Agriculture, 93(5), 11691175.
202
Cevallos-Cevallos, J. M., Reyes-De-Corcuera, J. I., Etxeberria, E., Danyluk, M.D., & Rodrick,
G. E. (2009). Metabolomic analysis in food science: A review. Trends in Food Science
& Technology, 20, 557566.
Chabreyrie, D., Chauvet, S., Guyon, F., Salagoty, M. H. U. C., Antinelli, J. F., & Medina, B.
(2008). Characterization and quantication of grape variety by means of shikimic
acid concentration and protein ngerprint in still white wines. Journal of Agricultural
Chen, L., Wang, J., Ye, Z., Zhao, J., Xue, X., Heyden, Y. V., et al. (2012). Classication of
Chinese honeys according to their oral origin by near infrared spectroscopy. Food
Chemistry, 135, 338342.
Christie, W. W. (2003). (3 ed.)Lipid analysis: Isolation, separation, identication and
lipidomic analysis, Vol. 15, Bridgewater: PJ Barnes & Associates.
Christoph, N., Rossmann, A., & Voerkelius, S. (2003). Possibilities and limitations of wine
authentication using stable isotope and meteorological data, data banks and statistical
tests. Part 1: Wines from Franconia and Lake Constance 1992 to 2001. Mitteilungen
Klosterneuburg, 53, 2340.
Cifuentes, A. (2009). Food analysis and foodomics. Journal of Chromatography A, 1216,
7109.
Cocchi, M., Durante, C., Marchetti, A., Armanino, C., & Casale, M. (2007). Characterization
and discrimination of different aged Aceto Balsamico Tradizionale di Modena
products by head space mass spectrometry and chemometrics. Analytica Chimica
Acta, 589, 96104.
Codex Alimentarius (2001). Codex standard for honey, CODEX STAN 12-1981.
Commission of the European communities (2002). Commission decision of 12 August
2002 implementing Council Directive 96/23/EC concerning the performance of
analytical methods and the interpretation of results (2002/657/EC). Ofcial Journal
of the European Communities, L 221, 836.
Concha-Herrera, V., Lerma-Garcia, M. J., Herrero-Martinez, J. M., & Simo-Alfonso, E. F.
(2009). Prediction of the genetic variety of extra virgin olive oils produced at La
Comunitat Valenciana, Spain, by Fourier transform infrared spectroscopy. Journal of
Agricultural and Food Chemistry, 57, 99859989.
Consonni, R., & Cagliani, L. R. (2008). Geographical characterization of polyoral and
acacia honeys by nuclear magnetic resonance and chemometrics. Journal of
Cozzolino, D. (2011). Infrared methods for high throughput screening of metabolites:
Food and medical applications. Combinatorial Chemistry & High Throughput Screening,
14, 125131.
Cozzolino, D., Cynkar, W. U., Shah, N., & Smith, P. A. (2011). Can spectroscopy
geographically classify Sauvignon Blanc wines from Australia and New Zealand?
Cuadros-Inostroza, A., Giavalisco, P., Hummel, J., Eckardt, A., Willmitzer, L., & Pena-Cortes,
H. (2010). Discrimination of wine attributes by metabolome analysis. Analytical
Chemistry, 82, 35733580.
De Alda-Garcilope, C., Gallego-Pic, A., Bravo-Yage, J. C., Garcinuno-Martnez, R. M., &
Fernndez-Hernando, P. (2012). Characterization of Spanish honeys with protected
designation of origin Miel de Granada according to their mineral content. Food
Chemistry, 135, 17851788.
de la Mata, P., Dominguez-Vidal, A., Bosque-Sendra, J. M., Ruiz-Medina, A.,
Cuadros-Rodrguez, L., & Ayora-Caada, M. J. (2012). Olive oil assessment in edible
oil blends by means of ATR-FTIR and chemometrics. Food Control, 23, 449455.
De Luca, M., Terouzi, W., Ioele, G., Kzaiber, F., Oussama, A., Oliverio, F., et al. (2011).
Derivative FTIR spectroscopy for cluster analysis and classication of Morocco olive
oils. Food Chemistry, 124, 11131118.
Dionisi, F., Prodolliet, J., & Tagliaferri, E. (1995). Assessment of olive oil adulteration by
reversed-phase high-performance liquid chromatography/amperometric detection
of tocopherols and tocotrienols. JAOCS, Journal of the American Oil Chemists' Society,
72, 15051511.
Donarski, J. A., Jones, S. A., & Charlton, A. J. (2008). Application of cryoprobe H-1 nuclear
magnetic resonance spectroscopy and multivariate analysis for the verication of
Corsican honey. Journal of Agricultural and Food Chemistry, 56, 54515456.
Drivelos, S. A., & Georgiou, C. A. (2012). Multi-element and multi-isotope-ratio analysis to
determine the geographical origin of foods in the European Union. TrAC Trends in
Analytical Chemistry, 40, 3851.
Dunn, W. B., Wilson, I. D., Nicholls, A. W., & Broadhurst, D. (2012). The importance of
experimental design and QC samples in large-scale and MS-driven untargeted
metabolomic studies of humans. Bioanalysis, 4, 22492264.
Etzold, E., & Lichtenberg-Kraag, B. (2008). Determination of the botanical origin of honey
by Fourier-transformed infrared spectroscopy: An approach for routine analysis.
European Food Research and Technology, 227, 579586.
European Commission (2013). Commission implementing regulation (EU) no 299/2013
of March 2013 amending regulation (EEC) no. 2568/91 on the characteristics of
olive oil and olive-residue oil and on the relevant methods of analysis. Ofcial Journal
of the European Union, L90, 5270.
European Community (1992). Regulation 2081/92. Ofcial Journal of the European Union,
L208.
Fauhl, C. (2006). Measurement uncertainty in wine appreciation. Mitteilungen
Klosterneuburg, 56, 313.
Fiehn Lab (2012). Peak alignment of LC, GC, MS, NMR data. http://ehnlab.ucdavis.edu/
staff/kind/Metabolomics/Peak_Alignment/
Fiehn, O., Robertson, D., Grifn, J., Werf, M., Nikolau, B., Morrison, N., et al. (2007). The
metabolomics standards initiative (MSI). Metabolomics, 3, 175178.
Frank, T., Scholz, B., Peter, S., & Engel, K. H. (2011). Metabolite proling of barley:
Inuence of the malting process. Food Chemistry, 124, 948957.
Fudge, A. L., Wilkinson, K. L., Ristic, R., & Cozzolino, D. (2013). Synchronous twodimensional MIR correlation spectroscopy (2D-COS) as a novel method for screening
smoke tainted wine. Food Chemistry, 139, 115119.
Garcia, J. S., Vaz, B. G., Corilo, Y. E., Ramires, C. F., Saraiva, S. R. A., Sanvido, G. B., et al. (2013).
Whisky analysis by electrospray ionization-Fourier transform mass spectrometry.
Food Research International, 51, 98106.
Garca-Canas, V. A., Sim, C., Herrero, M., Ibnez, E., & Cifuentes, A. (2012). Present and
future challenges in food analysis: Foodomics. Analytical Chemistry, 84(23),
1015010159.
Garrett, R., Schmidt, E. M., Pereira, L. F., Kitzberger, C. N. S. G., Scholz, M. B. G., Eberlin,
M. N., et al. (2013). Discrimination of arabica coffee cultivars by electrospray
ionization Fourier transform ion cyclotron resonance mass spectrometry and
chemometrics. LWT - Food Science and Technology, 50, 496502.
Gary, R. T., & Ebeler, Susan E. (2011). Progress in authentication of food and wine. Progress in
authentication of food and wine (pp. 311) (1081 ed.). : American Chemical Society.
Godelmann, R., Fang, F., Humpfer, E., Schtz, B., Bansbach, M., Schfer, H., et al. (2013).
Targeted and nontargeted wine analysis by 1H NMR spectroscopy combined with
multivariate statistical analysis. Differentiation of important parameters: Grape
variety, geographical origin, year of vintage. Journal of Agricultural and Food
Chemistry, 61, 56105619.
Goodacre, R., Broadhurst, D., Smilde, A., Kristal, B., Baker, J.D., Beger, R., et al. (2007).
Proposed minimum reporting standards for data analysis in metabolomics.
Metabolomics, 3, 231241.
Goodner, K. L., & Rouseff, R. L. (2001). Using an ion-trap MS sensor to differentiate and
identify individual components in grapefruit juice headspace volatiles. Journal of
Guillou, C. (2012). Traceable, readable, useable and expandable (TRUE) data in ofcial
food control (personal communication).
Guo, J., Yue, T., & Yuan, Y. (2012). Feature selection and recognition from nonspecic
volatile proles for discrimination of apple juices according to variety and geographical
origin. Journal of Food Science, 77, C1090C1096.
Hajslova, J., Cajka, T., & Vaclavik, L. (2011). Challenging applications offered by direct
analysis in real time (DART) in food-quality and safety analysis. TrAC Trends in
Analytical Chemistry, 30, 204218.
Hardy, N., & Taylor, C. (2007). A roadmap for the establishment of standard data exchange
structures for metabolomics. Metabolomics, 3, 243248.
Hendriks, M. M. W. B., Eeuwijk, F. A., Jellema, R. H., Westerhuis, J. A., Reijmers, T. H.,
Hoefsloot, H. C. J., et al. (2011). Data-processing strategies for metabolomics studies.
TrAC Trends in Analytical Chemistry, 30, 16851698.
Hennessy, S., Downey, G., & O'Donnell, C. (2008). Multivariate analysis of attenuated total
reection-Fourier transform infrared spectroscopic data to conrm the origin of
honeys. Applied Spectroscopy, 62, 11151123.
Hennessy, S., Downey, G., & O'Donnell, C. P. (2009). Conrmation of food origin claims by
Fourier transform infrared spectroscopy and chemometrics: Extra virgin olive oil
from Liguria. Journal of Agricultural and Food Chemistry, 57, 17351741.
Hennessy, S., Downey, G., & O'Donnell, C. P. (2010). Attempted conrmation of the
provenance of Corsican PDO honey using FT-IR spectroscopy and multivariate data
analysis. Journal of Agricultural and Food Chemistry, 58, 94019406.
Herrero, M., Garca-Caas, V., Simo, C., & Cifuentes, A. (2010). Recent advances in the
application of capillary electromigration methods for food analysis and foodomics.
Electrophoresis, 31, 205228.
Herrero, M., Sim, C., Garca-Canas, V., Ibnez, E., & Cifuentes, A. (2012). Foodomics:
MS-based strategies in modern food science and nutrition. Mass Spectrometry
Reviews, 31, 4969.
Imparato, G., Paolo, E. D., Braca, A., & Lamanna, R. (2011). Nuclear magnetic resonance
proling of wine blends. Journal of Agricultural and Food Chemistry, 59, 44294434.
Ioannou-Papayianni, E., Kokkinofta, R. I., & Theocharis, C. R. (2011). Authenticity of
Cypriot sweet wine commandaria using FT-IR and chemometrics. Journal of Food
Science, 76, C420C427.
Javidnia, K., Parish, M., Karimi, S., & Hemmateenejad, B. (2013). Discrimination of edible
oils and fats by combination of multivariate pattern recognition and FT-IR spectroscopy:
A comparative study between different modeling methods. Spectrochimica Acta Part A:
Molecular and Biomolecular Spectroscopy, 104, 175181.
Katajamaa, M., & Orei, M. (2007). Data processing for mass spectrometry-based
metabolomics. Journal of Chromatography A, 1158, 318328.
Kbler, H., Monakhova, Y. B., Kuballa, T., Tschiersch, C., Vancutsem, J., Thielert, G., et al.
(2011). Nuclear magnetic resonance spectroscopy and chemometrics to identify
pine nuts that cause taste disturbance. Journal of Agricultural and Food Chemistry,
59, 68776881.
Koda, M., Furihata, K., Wei, F., Miyakawa, T., & Tanokura, M. (2012). NMR-based metabolic
proling of rice wines by F 2-selective total correlation spectra. Journal of Agricultural
Koek, M. M., Jellema, R. H., van der Greef, J., Tas, A.C., & Hankemeier, T. (2011).
Quantitative metabolomics based on gas chromatography mass spectrometry: Status
and perspectives. Metabolomics, 7, 307328.
Kunz, M. R., Ottaway, J., Kalivas, J. H., Georgiou, C. A., & Mousdis, G. A. (2011). Updating a
synchronous uorescence spectroscopic virgin olive oil adulteration calibration
to a new geographical region. Journal of Agricultural and Food Chemistry, 59,
10511057.
Lachenmeier, D. W., Humpfer, E., Fang, F., Schutz, B., Dvortsak, P., Sproll, C., et al. (2009).
NMR-spectroscopy for nontargeted screening and simultaneous quantication of
health-relevant compounds in foods: The example of melamine. Journal of
Lai, Y. W., Kemsley, E. K., & Wilson, R. H. (1994). Potential of Fourier transform-infrared
spectroscopy for the authentication of vegetable-oils. Journal of Agricultural and
Laube, I., Hird, H., Brodmann, P., Ullmann, S., Sch+ne-Michling, M., Chisholm, J., et al.
(2010). Development of primer and probe sets for the detection of plant species in
honey. Food Chemistry, 118, 979986.

Le Gall, G., & Colquhoun, I. J. (2003). NMR spectroscopy in food authentication. In M. Lees
(Ed.), Food authenticity and traceability (pp. 131155). Cambridge: Woodhead
Publishing Limited.
Lees, M. (2003). Food authenticity and traceability. : Woodhead Pub.
Li Vigni, M., Durante, C., & Cocchi, M. (2013). Chapter 3 Exploratory data analysis. In M.
Federico (Ed.), Data handling in science and technology chemometrics in food chemistry,
Vol. 28. (pp. 55126): Elsevier.
Li, S., Shan, Y., Zhu, X., Zhang, X., & Ling, G. (2012). Detection of honey adulteration by
high fructose corn syrup and maltose syrup using Raman spectroscopy. Journal of
Food Composition and Analysis, 28, 6974.
Li-Chan, E., Chalmers, J. M., & Grifths, P. (2010). Applications of vibrational spectroscopy in
food science, Vol. Bd. 1, : John Wiley & Sons.
Lin, P., Chen, Y., & He, Y. (2012). Identication of geographical origin of olive oil using
visible and near-infrared spectroscopy technique combined with chemometrics.
Food and Bioprocess Technology, 5, 235242.
Liu, L., Cozzolino, D., Cynkar, W. U., Dambergs, R. G., Janik, L., O'Neill, B. K., et al. (2008).
Preliminary study on the application of visible-near infrared spectroscopy and
chemometrics to classify Riesling wines from different countries. Food Chemistry,
106, 781786.
Lolli, M., Bertelli, D., Plessi, M., Sabatini, A. G., & Restani, C. (2008). Classication of Italian
honeys by 2D HR-NMR. Journal of Agricultural and Food Chemistry, 56, 12981304.
Longobardi, F., Ventrella, A., Napoli, C., Humpfer, E., Schtz, B., Schfer, H., et al. (2012).
Classication of olive oils according to geographical origin by using 1H NMR
ngerprinting combined with multivariate analysis. Food Chemistry, 130, 177183.
Lopez-Rituerto, E., Savorani, F., Avenoza, A., Busto, J. S. H., Peregrina, J. S. M., & Engelsen, S.
R. B. (2012). Investigations of la Rioja terroir for wine production using 1H NMR
metabolomics. Journal of Agricultural and Food Chemistry, 60, 34523461.
Louw, L., Roux, K., Tredoux, A., Tomic, O., Naes, T., Nieuwoudt, H. H., et al. (2009).
Characterization of selected South African young cultivar wines using FTMIR
spectroscopy, gas chromatography, and multivariate data analysis. Journal of
Mannina, L., D'Imperio, M., Capitani, D., Rezzi, S., Guillou, C., Mavromoustakos, T., et al.
(2009). 1H NMR-based protocol for the detection of adulterations of rened olive
oil with rened hazelnut oil. Journal of Agricultural and Food Chemistry, 57, 1155011556.
Mannina, L., Marini, F., Gobbino, M., Sobolev, A. P., & Capitani, D. (2010). NMR and
chemometrics in tracing European olive oils: The case study of Ligurian samples.
Talanta, 80, 21412148.
Martelo-Vidal, M. J., Domnguez-Agis, F., & Vzquez, M. (2013). Ultraviolet/visible/
near-infrared spectral analysis and chemometric tools for the discrimination of
wines between subzones inside a controlled designation of origin: A case study of
R + as Baixas. Australian Journal of Grape and Wine Research, 19, 6267.
Martnez-Lozano Sinues, P., Alonso-Salces, R. M., Zingaro, L., Finiguerra, A., Holland, M. V.,
Guillou, C., et al. (2012). Mass spectrometry ngerprinting coupled to National
Institute of Standards and Technology Mass Spectral search algorithm for pattern
recognition. Analytica Chimica Acta, 755, 2836.
Matos, J. T. V., Duarte, R. M. B. O., & Duarte, A.C. (2012). Trends in data processing of
comprehensive two-dimensional chromatography: State of the art. Journal of
Chromatography, B: Analytical Technologies in the Biomedical and Life Sciences, 910,
3145.
Mazzei, P., Francesca, N., Moschetti, G., & Piccolo, A. (2010). NMR spectroscopy evaluation
of direct relationship between soils and molecular composition of red wines from
Aglianico grapes. Analytica Chimica Acta, 673, 167172.
Mignani, A. G., Ciaccheri, L., Ottevaere, H., Thienpont, H., Conte, L., Marega, M., et al.
(2011). Visible and near-infrared absorption spectroscopy by an integrating sphere
and optical bers for quantifying and discriminating the adulteration of extra virgin
olive oil from Tuscany. Analytical and Bioanalytical Chemistry, 399, 13151324.
Monakhova, Y. B., Schfer, H., Humpfer, E., Spraul, M., Kuballa, T., & Lachenmeier, D. W.
(2011). Application of automated eightfold suppression of water and ethanol signals
in 1H NMR to provide sensitivity for analyzing alcoholic beverages. Magnetic
Resonance in Chemistry, 49, 734739.
Moore, J. C., Spink, J., & Lipp, M. (2012). Development and application of a database of
food ingredient fraud and economically motivated adulteration from 1980 to 2010.
Journal of Food Science, 77, R118R126.
Nietner, T., Pster, M., Glomb, M.A., & Fauhl-Hassek, C. (2013). Authentication of the
botanical and geographical origin of distillers dried grains and solubles (DDGS) by
FT-IR spectroscopy. Journal of Agricultural and Food Chemistry, 61, 72257233.
Obaidat, S. M., Kahnfar, M. S., & Obaidat, W. M. (2009). Classication of edible oils and
uncovering adulteration of virgin olive oil using FTIR with the aid of chemometrics.
Australian Journal of Basic and Applied Sciences, 3, 20482053.
Ogrinc, N., Kosir, I. J., & Spangenberg, J. (2003). The application of NMR and MS methods
for detection of adulteration of wine, fruit juices, and olive oil. A review. Analytical
and Bioanalytical Chemistry, 376, 424430.
Oliver, S. G., Winson, M. K., Kell, D. B., & Baganz, F. (1998). Systematic functional analysis
of the yeast genome. Trends in Biotechnology, 16, 373378.
Oliveri, P., Di Egidio, V., Woodcock, T., & Downey, G. (2011). Application of classmodelling techniques to near infrared data for food authentication purposes. Food
Chemistry, 125, 14501456.
Olivieri, A.C. (2008). Analytical advantages of multivariate data processing. One, two,
three, innity? Analytical Chemistry, 80, 57135720.
MoniQA (2013). Food authenticity General considerations. http://www.moniqa.eu/
authenticity.16-9-2013
Ourrach, I., Rada, M., Prez-Camino, M. C., Benaissa, M., & Guinda, . (2012). Detection of
argan oil adulterated with vegetable oils: New markers. Grasas y Aceites, 63, 355364.
Oussama, A., Elabadi, F., Platikanov, S., Kzaiber, F., & Tauler, R. (2012). Detection of olive oil
adulteration using FT-IR spectroscopy and PLS with variable importance of projection
(VIP) scores. JAOCS, Journal of the American Oil Chemists' Society, 89, 18071812.
203
Papotti, G., Bertelli, D., Graziosi, R., Silvestri, M., Bertacchini, L., Durante, C., et al. (2013).
Application of one- and two-dimensional NMR spectroscopy for the characterization
of protected designation of origin Lambrusco wines of Modena. Journal of Agricultural
Pierce, K. M., Kehimkar, B., Marney, L. C., Hoggard, J. C., & Synovec, R. E. (2012). Review of
chemometric analysis techniques for comprehensive two dimensional separations
data. Journal of Chromatography A, 1255, 311.
Pierna, J. A. F., Abbas, O., Dardenne, P., & Baeten, V. (2011). Discrimination of Corsican
honey by FT-Raman spectroscopy and chemometrics. Biotechnologie, Agronomie,
Socit et Environnement, 15, 7584.
Pierna, J. A. F., Duponchel, L., Ruckebusch, C., Bertrand, D., Baeten, V., & Dardenne, P.
(2012). Trappist beer identication by vibrational spectroscopy: A chemometric
challenge posed at the Chimiomtrie 2010 congress. Chemometrics and Intelligent
Laboratory Systems, 113, 29.
Pizarro, C., Rodrguez-Tecedor, S., Prez-del-Notario, N., Esteban-Dez, I., & Gonzlez-Siz,
J. M. (2013). Classication of Spanish extra virgin olive oils by data fusion of visible
spectroscopic ngerprints and chemical descriptors. Food Chemistry, 138, 915922.
Pizarro, C., Rodriguez-Tecedor, S., Perez-del-Notario, N., & Gonzalez-Saiz, J. (2011).
Recognition of volatile compounds as markers in geographical discrimination of
Spanish extra virgin olive oils by chemometric analysis of non-specic chromatography
volatile proles. Journal of Chromatography, A, 1218, 518523.
Primrose, S., Woolfe, M., & Rollinson, S. (2010). Food forensics: Methods for determining
the authenticity of foodstuffs. Trends in Food Science & Technology, 21, 582590.
Riovanto, R., Cynkar, W. U., Berzaghi, P., & Cozzolino, D. (2011). Discrimination between
Shiraz wines from different Australian regions: The role of spectroscopy and
chemometrics. Journal of Agricultural and Food Chemistry, 59, 1035610360.
Rodriguez-Saona, L. E., & Allendorf, M. E. (2011). Use of FTIR for rapid authentication and
detection of adulteration of food. Annual Review of Food Science and Technology, 2,
467483.
Roessner, U., Nahid, A., Chapman, B., Hunter, A., & Bellgrad, M. (2011). Metabolomics
The combination of analytical biochemistry, biology, and informatics. Comprehensive
Biotechnology (pp. 447459) (second ed.). : Elsevier.
Rohman, A., & Che Man, Y. B. (2011). Simultaneous quantitative analysis of two functional
food oils, extra virgin olive oil and virgin coconut oil using FTIR spectroscopy and
multivariate calibration. International Food Research Journal, 18, 12311235.
Rohman, A., & Che Man, Y. B. (2012). Quantication and classication of corn and
sunower oils as adulterants in olive oil using chemometrics and FTIR spectra. The
Scientic World Journal, 2012.
Rohman, A., Che Man, Y. B., Hashim, P., & Ismail, A. (2011). FTIR spectroscopy combined
with chemometrics for analysis of lard adulteration in some vegetable oils. CyTA
Journal of Food, 9, 96101.
Rohman, A., Che Man, Y. B., Ismail, A., & Hashim, P. (2010). Application of FTIR spectroscopy
for the determination of virgin coconut oil in binary mixtures with olive oil and
palm oil. JAOCS, Journal of the American Oil Chemists' Society, 87, 601606.
Rohman, A., & Man, Y. B. C. (2010). Fourier transform infrared (FTIR) spectroscopy for
analysis of extra virgin olive oil adulterated with palm oil. Food Research International,
43, 886892.
Rohman, A., & Man, Y. B. C. (2011). Potential use of FTIR-ATR spectroscopic method for
determination of virgin coconut oil and extra virgin olive oil in ternary mixture
systems. Food Analytical Methods, 4, 155162.
Rohman, A., & Man, Y. B. C. (2012a). Application of Fourier transform infrared
spectroscopy for authentication of functional food oils. Applied Spectroscopy
Reviews, 47, 113.
Rohman, A., & Man, Y. B. C. (2012b). The chemometrics approach applied to FTIR spectral
data for the analysis of rice bran oil in extra virgin olive oil. Chemometrics and
Intelligent Laboratory Systems, 110, 129134.
Ruiz-Sambls, C., Tres, A., Koot, A., Van Ruth, S. M., Gonzlez-Casado, A., & CuadrosRodrguez, L. (2012). Proton transfer reactionmass spectrometry volatile organic
compound ngerprinting for monovarietal extra virgin olive oil identication. Food
Chemistry, 134, 589596.
Sangster, T., Major, H., Plumb, R., Wilson, A. J., & Wilson, I. D. (2006). A pragmatic and
readily implemented quality control strategy for HPLCMS and GCMS-based
metabonomic analysis. Analyst, 131, 10751078.
Schievano, E., Peggion, E., & Mammi, S. (2010). H-1 Nuclear magnetic resonance spectra
of chloroform extracts of honey for chemometric determination of its botanical
origin. Journal of Agricultural and Food Chemistry, 58, 5765.
Schievano, E., Stocchero, M., Morelato, E., Facchin, C., & Mammi, S. (2012). An NMR-based
metabolomic approach to identify the botanical origin of honey. Metabolomics, 8,
679690.
SGF International e.V. (2012). SGF Proling. http://www.sgf.org/en/home/forschungund-projekte/proling/
Shen, F., Yang, D., Ying, Y., Li, B., Zheng, Y., & Jiang, T. (2012). Discrimination between
Shaoxing wines and other Chinese rice wines by near-infrared spectroscopy and
chemometrics. Food and Bioprocess Technology, 5, 786795.
Smilde, A., Bro, R., & Geladi, P. (2004). Algorithms. Multi-way analysis with applications in
the chemical sciences (pp. 111144). : John Wiley & Sons, Ltd.
Smolinska, A., Blanchet, L., Buydens, L. M. C., & Wijmenga, S. S. (2012). NMR and pattern
recognition methods in metabolomics: From data acquisition to biomarker discovery:
A review. Analytica Chimica Acta, 750, 8297.
Smyth, H. E., Cozzolino, D., Cynkar, W. U., Dambergs, R. G., Sefton, M., & Gishen, M. (2008).
Near infrared spectroscopy as a rapid tool to measure volatile aroma compounds in
Riesling wine: Possibilities and limits. Analytical and Bioanalytical Chemistry, 390,
19111916.
Son, H. S., Hwang, G. S., Ahn, H. J., Park, W. M., Lee, C. H., & Hong, Y. S. (2009).
Characterization of wines from grape varieties through multivariate statistical
analysis of H-1 NMR spectroscopic data. Food Research International, 42, 14831491.
204
Son, H. S., Hwang, G. S., Kim, K. M., Ahn, H. J., Park, W. M., van den Berg, F., et al. (2009).
Metabolomic studies on geographical grapes and their wines using 1H NMR analysis
coupled with multivariate statistics. Journal of Agricultural and Food Chemistry, 57,
14811490.
Spraul, M., Schutz, B., Rinke, P., Koswig, S., Humpfer, E., Schafer, H., et al. (2009). NMR-based
multi parametric quality control of fruit juices: SGF proling. Nutrients, 1, 148155.
Teletchea, F., Maudet, C., & Hnni, C. (2005). Food and forensic molecular identication:
Update and challenges. Trends in Biotechnology, 23, 359366.
Theodoridis, G. A., Gika, H. G., Want, E. J., & Wilson, I. D. (2012). Liquid chromatography
mass spectrometry based global metabolite proling: A review. Analytica Chimica
Acta, 711, 716.
TRACE (2013). TRACE Tracing the origin of food. http://www.trace.eu.org/index.php
(Retrieved 2013)
Trygg, J., Gullberg, J., Johansson, A. I., Jonsson, P., & Moritz, T. (2006). Chemometrics in
metabolomics An introduction. In K. Saito, R. Dixon, & L. Willmitzer (Eds.), Plant
Metabolomics (pp. 117128) (57 ed.). Berlin Heidelberg: Springer.
Trygg, J., Holmes, E., & Lundstedt, T. (2006). Chemometrics in metabonomics. Journal of
Proteome Research, 6, 469479.
Vaclavik, L., Cajka, T., Hrbek, V., & Hajslova, J. (2009). Ambient mass spectrometry
employing direct analysis in real time (DART) ion source for olive oil quality and
authenticity assessment. Analytica Chimica Acta, 645, 5663.
Vaclavik, L., Lacina, O., Hajslova, J., & Zweigenbaum, J. (2011). The use of high performance
liquid chromatographyquadrupole time-of-ight mass spectrometry coupled to
advanced data mining and chemometric tools for discrimination and classication
of red wines according to their variety. Analytica Chimica Acta, 685, 4551.
Van, Q. N., Issaq, H. J., Jiang, Q., Li, Q., Muschik, G. M., Waybright, T. J., et al. (2007).
Comparison of 1D and 2D NMR spectroscopy for metabolic proling. Journal of
Proteome Research, 7, 630639.
Varmuza, K., & Filzmoser, P. (2009). Introduction to multivariate statistical analysis in
chemometrics. : Taylor & Francis.
Villagra, E., Santos, L. S., Vaz, B. G., Eberlin, M. N., & Felipe Laurie, V. (2012). Varietal
discrimination of Chilean wines by direct injection mass spectrometry analysis
combined with multivariate statistics. Food Chemistry, 131, 692697.
Vinci, G., Preti, R., Tieri, A., & Vieri, S. (2013). Authenticity and quality of animal origin food
investigated by stable-isotope ratio analysis. Journal of the Science of Food and
Agriculture, 93, 439448.
Von Baer, D., Mardones, C., Guti+rrez, L., Hofmann, G., Hitschfeld, A., & Vergara, C.
(2007). Anthocyanin, avonol, and shikimic acid proles as a tool to verify varietal
authenticity in red wines produced in Chile, 952, 228238.
Waiblinger, H. U., Ohmenhaeuser, M., Meissner, S., Schillinger, M., Pietsch, K., Goerlich, O.,
et al. (2012). In-house and interlaboratory validation of a method for the extraction
of DNA from pollen in honey. Journal fr Verbraucherschutz und Lebensmittelsicherheit,
7, 243254.
Ward, J., Baker, J., Miller, S., Deborde, C., Maucourt, M., Biais, B., et al. (2010). An
inter-laboratory comparison demonstrates that [1H]-NMR metabolite ngerprinting
is a robust technique for collaborative plant metabolomic data collection.
Metabolomics, 6, 263273.
Wei, F., Furihata, K., Koda, M., Hu, F., Kato, R., Miyakawa, T., et al. (2012). 13C NMR-based
metabolomics for the classication of green coffee beans according to variety and
origin. Journal of Agricultural and Food Chemistry, 60, 1011810125.
Westad, F., Bevilacqua, M., & Marini, F. (2013). Chapter 4 Regression. In M. Federico
(Ed.), Data handling in science and technology chemometrics in food chemistry, Vol.
28. (pp. 127170): Elsevier.
Woodcock, T., Downey, G., & O'Donnell, C. P. (2008). Conrmation of declared provenance
of European extra virgin olive oil samples by NIR spectroscopy. Journal of Agricultural
Woodcock, T., Downey, G., & O'Donnell, C. P. (2009). Near infrared spectral
ngerprinting for conrmation of claimed PDO provenance of honey. Food Chemistry,
114, 742746.
Woolfe, M., & Primrose, S. (2004). Food forensics: Using DNA technology to combat
misdescription and fraud. Trends in Biotechnology, 22, 222226.
Yoshida, H., Yamazaki, J., Ozawa, S., Mizukoshi, T., & Miyano, H. (2009). Advantage of
LCMS metabolomics methodology targeting hydrophilic compounds in the
studies of fermented food samples. Journal of Agricultural and Food Chemistry,
57, 11191126.
Zhang, Y. L., Chen, J. B., Lei, Y., Zhou, Q., Sun, S. Q., & Noda, I. (2010). Discrimination
of different red wine by Fourier-transform infrared and two-dimensional
infrared correlation spectroscopy. Journal of Molecular Structure, 974,
144150.
Zhang, J., Li, L., Gao, N., Wang, D., Gao, Q., & Jiang, S. (2010). Feature extraction and
selection from volatile compounds for analytical classication of Chinese red wines
from different varieties. Analytica Chimica Acta, 662, 137142.
Zhu, X., Li, S., Shan, Y., Zhang, Z., Li, G., Su, D., et al. (2010). Detection of adulterants such as
sweeteners materials in honey using near-infrared spectroscopy and chemometrics.
Journal of Food Engineering, 101, 9297.
Zou, M. Q., Zhang, X. F., Qi, X. H., Ma, H. L., Dong, Y., Liu, C. W., et al. (2009). Rapid
authentication of olive oil adulteration by Raman spectrometry. Journal of Agricultural

Potential and Limitation On Authentication Food

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Potential and Limitation On Authentication Food

Uploaded by

Copyright:

Available Formats

Food Research International 60 (2014) 189204

Contents lists available at ScienceDirect

Food Research International

Potential and limitations of non-targeted ngerprinting for

S. Esslinger et al. / Food Research International 60 (2014) 189204

the corresponding denitions in metabolomics (Table 1, Koek, Jellema,

Aim and advantage/disadvantage

Aim: e.g. authentication/characterization of food (Fauhl, 2006; Koek et al., 2011)

Using omics techniques e.g.

Aim: e.g. comprehensive authentication/

(Antignac et al., 2011; Cevallos-Cevallos et al., 2009; Koek et al., 2011)

(Cifuentes, 2009; Garca-Canas, Sim, Herrero, Ibnez, & Cifuentes,

S. Esslinger et al. / Food Research International 60 (2014) 189204

of sample or its extract in a non-destructive, easy, quick and direct (with

of sample material(s) and analytical technique(s) which will be

establishment & sampling

e.g. by the use of databases

by the use of supervised

Fig. 1. The procedure/strategy of ngerprinting techniques for food analysis.

S. Esslinger et al. / Food Research International 60 (2014) 189204

system coupled to a ame ionization detector (GC-FID), followed by

spectroscopy were shown to be most useful tools in non-targeted

S. Esslinger et al. / Food Research International 60 (2014) 189204

overlapping signals. A major drawback might be the loss of a considerable

Stable isotope analysis

Stable isotope analysis

Stable isotope analysis

Stable isotope analysis

Stable isotope analysis (13C)

S. Esslinger et al. / Food Research International 60 (2014) 189204

PCA, HCA, KNN, (Boffo, Tavares, Tobias, Ferreira, &

Extraction and clean-up

Extract with chloroform, pH

Dilution with distilled water,

OPLS-DA, OPLSHCA, SIMCA

1st, 2nd derivative, SNV

SNV, 1st, 2nd derivative

Homogenization, dilution with

Homogenization, dilution with

None, different derivatives,

SNV, 1st, 2nd derivative

(Cavazza, Corradini, Musci &

(Etzold & Lichtenberg-Kraag,

(Aliferis, Tarantilis, Harizanis, &

(Hennessy, Downey, & O'Donnell,

(Consonni & Cagliani, 2008)

(Donarski et al., 2008)

(Hennessy et al., 2008)

(Javidnia, Parish, Karimi, &

SNV, 1st derivative

PCA, LDA, PLSDA

Baseline correction, MSC,

Normalization, meancentering, weighting

1st, 2nd Derivative, SNV

Direct analysis (HS-SPME)

(Li et al., 2012)

(Ruiz-Sambls et al., 2012)

(Casale et al., 2012)

(Longobardi et al., 2012)

(Lin, Chen, & He, 2012)

(Martnez-Lozano Sinues et al.,

S. Esslinger et al. / Food Research International 60 (2014) 189204