You are on page 1of 6

Food Research International 63 (2014) 126131

Contents lists available at ScienceDirect

Food Research International

journal homepage:

NMR screening in the quality control of food and nutraceuticals

Anna Paola Minoja , Claudia Napoli
Bruker Italia S.r.l., Viale Lancetti 43, 20156 Milan, Italy

a r t i c l e

i n f o

Article history:
Received 3 November 2013
Received in revised form 11 April 2014
Accepted 16 April 2014
Available online 2 May 2014
Fruit Juice
Complex mixture screening

a b s t r a c t
Modern research in food science and nutrition requires innovative strategies using advanced analytical analysis.
Nuclear Magnetic Resonance (NMR) has proven to be a robust analytical tool for the screening and analysis of
complex mixtures such as body uids, food and beverages. The next logical application of NMR based screening
methods is for food quality control. Here we show a fully automated quality control analysis of fruit juice and
wine. This analysis requires the user to have minimal working knowledge of NMR and delivers reports with results a person untrained in NMR analysis can easily understand. However, to produce reliable results, substantial
effort in validation of quantication and statistical analysis is needed initially. These efforts are rewarded with
rapid highly reproducible screening capabilities. Further application of the same principles is under investigation
for nutraceuticals. Very promising results obtained to date suggest that NMR will provide a needed new analytical
tool for product assessment of these highly complex materials.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Traditionally NMR is recognized as a fundamental analytical tool
in structure verication, elucidation and purity analysis. It has been
widely applied in synthetic organic chemistry to investigate the molecular structure and to study kinetics, dynamics, and interactions
between molecules and protein conformations (Derome, 1987;
Keeler, 2005; Saunders & Hunter, 1993). Resulting from research in
the eld of Metabonomics (Lindon, Nicholson, Holmes, & Everett, 2000;
Nicholson, Lindon, & Holmes, 1999), NMR applications have expanded
into new areas that utilize its capabilities in screening and analysis of
complex mixtures. Initially established for body uid investigation,
NMR applications in Metabonomics were successfully extended to
other complex mixture samples, especially food stuffs (for a recent
review see Mannina, Viel, & Sobolev, 2012). Technological and methodological challenges for these new application areas require the highest
degree of automation at each step of the process: sample preparation,
acquisition, processing, analysis and reporting.
So far, NMR was mainly used in Food Research. In this article, we
focus our attention on the next phase of NMR developments in this application eld: NMR as a quality control tool for food material and the
now popular nutraceuticals. Quality control of food means control of
origin, properties, composition, safety, and authenticity and includes
all steps of the global supply chain; from the raw material up to the
nal product sold on the supermarket shelf. Many constituents are
present in the crude products whereas other components arise during

Corresponding author. Tel.: +39 0270636370; fax: +39 022361294.

E-mail address: (A.P. Minoja).
0963-9969/ 2014 Elsevier Ltd. All rights reserved.

storage and processing or can be added for preservation or to address

a specic nutritional concern. Consequently, food authenticity, composition and traceability are of great concern to consumers, food manufactures, retailer and regulatory bodies and analytical providers. Despite
the great concern these products are currently monitored and certied
by means of paper documentation, some targeted analysis and sensory
properties. Sensory properties are usually tested by human beings and
results depend on individual's perception that is inuenced by factors
like nutritional education, age and current trends. Sensory tests are
expensive, time consuming and cannot be applied to food that contains,
or could contain, poison or toxins. In principle, several analytical techniques can overcome most of these limitations, but in order to be
applied in the real world the following requirements must be fullled:
Reproducibility (site to site, country to country, instrument to
Targeted AND non-targeted (recognize even unknown deviations),
Safe compound identication,
Ability to identify known adulterants,
Absolute quantication (one calibration for all compounds),
Rapid and push-button, automatic reports,
Low cost per sample,
Low cost per parameter.
Because there is not just one single analytical technique to provide
the best evaluation for each of the above-mentioned requirements on
a complex mixture, the use of complementary techniques capable of
providing a broad range of information is desirable. Despite the lower
sensitivity in comparison with other analytical techniques, like MS,

A.P. Minoja, C. Napoli / Food Research International 63 (2014) 126131

and the signals' overlapping in the spectra, NMR analysis has a great advantage of being able to give information on all components present in
the sample, simultaneously. The single NMR spectrum contains all the
signals arising from all the protons of all the molecules present in the
sample. This, combined with the fact that the technique is highly
reproducible and inherently quantitative, allows the analyst to quickly
obtain information regarding the entire chemical composition of the
2. Experimental
For fruit juice sample preparation, NMR analysis and data analysis
see Spraul et al. (2009).
For wine sample preparation, NMR analysis and data analysis see
Monakhova et al. (2011).
For blueberry leaf sample preparation, NMR analysis and data analysis
see Hicks et al. (2012).
For Ginseng sample preparation, NMR analysis and data analysis see
Yuk et al. (2013).
3. Results and discussion
3.1. Fruit Juice quality control by NMR
A complex mixture contains many metabolites ranging widely in
concentration and chemical properties. For this reason, the 1H NMR
spectra can be highly complex due to the large number of signals and
overlapping resonances. The analysis of these samples requires appropriate strategies, namely targeted and non-targeted analyses. Specic
components are identied and consequently quantied by targeted
analysis, whereas non-targeted analysis detects unexpected deviations
by applying statistical models to the whole spectrum without the need
for interpretation of specic signals.
Spin Generated Fingerprint Proling (SGF-proling) (Spraul et al.,
2009) is a method for fruit juice analysis developed jointly by Bruker
BioSpin GmbH and SGF International e.V. With this example, the application of the principles established for metabonomics NMR as applied to
food quality control is demonstrated.
The rst analysis step is the targeted analysis to assess the analyzed
food matrix on the basis of specic deviations in concentrations of single
compounds that may indicate quality problems. A list of compounds can
be identied and quantied from a single NMR spectrum. The concentrations of specic components are very important because they allow
detecting quality problems like, for example, the addition of sugars.
Due to the direct proportionality of the integral of the NMR peaks and
the number of observed nuclei, NMR is an inherently quantitative technique and thus it is an efcient method for concentration determination. With appropriate selection of experiments and procedures, the
NMR does not need daily prior calibration. The determination of the
absolute concentration only requires an external reference compound
of known concentration. The absolute intensities of reference and unknown spectra are correlated by the measurement of a precise 360
pulse, direct consequence of the principle of reciprocity (Wider &
Dreier, 2006). In addition, the capability of NMR to quantify multiple
compounds in a mixture, from just one measurement, is a clear advantage
over the targeted standard analytical routines, where prior separation of
components is often needed, as it allows detection of the appearance of
unexpected ingredients and consequently of unknown frauds. The SGF
method is constantly being updated; currently concentrations of more
than 30 different compounds are determined (depending on the type of
juice). Furthermore, meaningful internal relations of these values are calculated, such as the ratio of glucose and fructose or the ratio of sucrose
versus total sugar.
The second step is the non-targeted analysis. SGF-proling includes
statistical routines to classify and/or verify samples, based on normality


models. In this second part of fruit juice analysis, the whole NMR spectrum is used as a ngerprint where every signal is taken into account regardless of whether it was used in the targeted analysis or not. Following
a cascade of classication models, each sample is classied and then validated with respect to the most highly probable group. The normality
model is a spectral database of more than 3000 samples with more
than 30 different types of fruits from more than 50 countries. At least
3040 samples of each subclass were used to build the models. The samples are fully authentic since they have been collected under the control
of an inspector from running production on site.
For the model building, a preparation step is needed in order to reduce the number of data points; this is done via bucketing in which
the spectra are divided in equally sized regions (buckets). Then, the
applied statistical method is a combination of Principal Component
Analysis (PCA) and Discriminant Analysis. The principle of PCA is to
get insight into trends (e.g. a separation into groups) in ensembles of
spectra, which would not be obvious in the usual spectroscopic space.
The variables, which span the spectroscopic space, are the buckets.
PCA involves a coordinate transform from the spectroscopic space into
a space, which is spanned by new variables, called Principal Components, which in turn are related to directions of largest variances in
the ensemble. The rst principal component explains most of the variance, the second component the second most, and so on. If interesting
features (groups, outliers) are observed in the score plots, the relation
between the variables in the new principal component space and original spectroscopic space, through the use of a loadings plot, allows understanding what the reason for this was. The loadings plot shows the
buckets, or chemical shift regions of the NMR spectrum, of highest variance. Discriminant Analysis (DA) is then performed, which searches for
linear combinations of variables that reect the largest possible separation between groups of interest. The statistical space built during PCA is
the new input matrix for DA, which leads to dene a further subspace
with optimum group separation. The obtained model is then validated
in order to dene the predictive capability.
The spectrum obtained from a newly analyzed sample, undergoes
the same statistical treatment, and the resulting variables are then compared with the normality model in order to detect similarities or deviations, by using statistical classication routines. The classication tools
include the estimation of: 1) the type of fruit (the genus, for example
Citrus), 2) the variety (for example orange), 3) the product type
(discriminating between direct juice and diluted from concentrate
juice) and 4) the geographical origin. After a sample has been classied
as belonging to the most accurate group, it is veried by comparing the
spectrum with a reference dataset (univariate analysis) in order to
detect deviations of metabolite concentrations and then with multivariate analysis, taking into account also the relation between signals deriving
from different molecules and corresponding to the expected ratios between the mixture components, in order to detect eventually deviations
that univariate does not show. Moreover, retrospective use of older data
in new statistical models or quantication is always possible as all the
H nuclei are acquired,
Results from both the targeted and non-targeted approaches are
obtainable in automation and do not require NMR interpretation from
the classical NMR spectrum. Examples of the quality control reports
which describe and classify the sample are shown in Figs. 1 and 2. Quantication results are listed in a table showing the concentration of every
metabolite identied (Fig. 1a). Where reference values exist for expected concentrations dened by the European Fruit Juice Association
(A.I.J.N.) for the same type of juice, the result is also expressed as a trafc
light (Flag), green if the obtained value is in the optimal range and red if
outside the optimal range. Moreover, the table shows, in the last column,
the distribution of metabolite concentration in the authentic set of
samples used for statistical model building described previously.
Multivariate classication results are shown as 3D plots where it is
possible to see, as ellipsoids, the statistical space occupied by every
authentic group, and the position of the new sample being measured


A.P. Minoja, C. Napoli / Food Research International 63 (2014) 126131

Fig. 1. Extract from the SGF nal report of an Orange Juice sample. a) Quantication results table shows: concentration, the comparison to the European Fruit Juice Association (A.I.J.N.)
values by trafc-light code, and the distribution of concentrations in the corresponding authentic model. b) Classication model: a specic sample is assigned to its most probable
group. The group is chosen from a list of proposed groups. The 3D-discrimination diagram shows the ellipsoids of available groups in the projection space of the NMR-proles with maximized discrimination. The star represents the actual sample.

(Fig. 1b). The mathematically correct probability for any group

membership is represented by p-values which are calculated in the
complete space. The univariate method of evaluating the sample,
the quantile plot (Fig. 2), shows the distribution of peak intensities
over the whole NMR spectrum. This allows an immediate comparison
between the actual sample and the whole reference dataset: a complex
spectrum with different colors representing the distribution of normal
intensities in the model. This is used for the verication of the classied
spectra and for detection of outliers because it is easy to see if there are
peaks lower or higher than the expected.

3.2. Wine quality control by NMR

SGF-proling is now a recognized method and continues to grow by
continual updating of the reference juice database and statistical
models. The same successful methodology is applied to other foodstuffs,
taking into account the specic properties of the food matrix itself. In
comparison to the protocol optimized for fruit juice, the investigation
of wine required modication both in sample preparation and in NMR
experiment acquisition. In wine samples a slight pH variation of the
order of magnitude of 0.005 units can change chemical shifts of NMR

Fig. 2. Quantile plot: Verication models are non-targeted analyses comparing the whole NMR-prole of a specic sample with one corresponding group of reference spectra from the
database. All spectra data points are taken into account irrespective of whether the signals are caused by already identied molecules or not.

A.P. Minoja, C. Napoli / Food Research International 63 (2014) 126131

signals. Such pH change is especially evident in the chemical shift of

metabolites with acidic or basic functional groups and interferes with
multivariate analysis. Consequently, the pH of the NMR sample is
adjusted by adding buffer solution and a small amount of acid or base
to precisely correct the pH. This critical step is also performed in automation and done directly to the small volume used for the NMR tube.
Generally speaking, in order to fully characterize food stuffs, it is important to investigate the role played by minor components whose concentrations can be of several orders of magnitude lower than the major
ones, but responsible for nutritional, tasting and safety properties. For
example, fruit juice is mainly composed of water and water is additionally added to the NMR sample as a lock solvent in the form of D2O. Consequently, the large water signal observed in the proton spectrum is
selectively suppressed by means of a presaturation pulse in the NMR
experiment in order to observe the signals of interest (see Spraul
et al., 2009). The need to eliminate major signals is even more important
when the investigated foodstuff is composed of more than one major
component and/or when NMR major component spectrum shows
more than one NMR signal. This is clearly the case of wine and, more
generally, of alcoholic beverages, (Monakhova, Kuballa, & Lachenmeier,
2012) which consist mainly of ethanol and water. In order to overcome
such a problem the choice is between the elimination of water and
ethanol during sample preparation and the suppression of water and
ethanol signals during the NMR experiment. Both approaches have
pros and cons, but because reproducibility and automation are the
main requirements that any quality control methodology must fulll,
sample manipulation is typically avoided. As a result, a methodology
was optimized to achieve an NMR spectrum with water and ethanol
signals suppressed without losing signals outside those regions and
still fullling the quality control requirements.
Therefore, the procedure for wine analysis involves rst acquiring an
NMR spectrum suppressing only the signal of OH protons with water


presaturation. This spectrum serves as the reference spectrum and provides the exact frequencies of the eight signals due to ethanol and water
in the sample (the OH singlet of water and ethanol, as well as the CH2
quartet and CH3 triplet of ethanol). Such frequencies are used to generate a shaped pulse that is applied during the relaxation delay to simultaneously suppress the water and ethanol signals without interfering
with signals from other molecules resonating at close frequencies
(Monakhova et al., 2011). The combination of these experiments, run
under full automation, allows the application of the same analysis
scheme optimized for fruit juice even for wine analysis, including components' quantication, sample classication and sample verication. The
methodology, tested and patented under the name Winescreener,
provides the result as a report, an excerpt of which is shown in Fig. 3.

3.3. NMR and nutraceuticals

One of the most important aspects of the analytical characterization
of food is related to the relative health consequences, which can be both
harmful and healthy. The connection between food and health is so important that food, or parts of food, that provide medical or health benets
are dened by the term nutraceuticals (Zeisel, 1999). Due to the pharmacological, or at least wellness properties declared, nutraceuticals should
be certied according to even stricter criteria than those applied to
food. In principle, the quality standard certication should be at the pharmaceutical level. Actually, the present situation is rather confused even
in dening nutraceuticals themselves. The concepts of nutraceuticals,
functional foods, dietary supplements, and phytochemicals are often
used interchangeably. The investigation of how methodologies and technologies successfully applied to both food and medicinal agents can help
in dening innovative quality control protocols is urgently needed
(Wrick, 2005).

Fig. 3. Extract from Winescreener report of a Sauvignon Blanc sample. a) Result summary: a trafc light indicates if classication, quantication, and verication are consistent for the
actual sample. b) Classication model for the wine type assignment, representing the probability of classication for every group. c) Quantication results table: shows concentration, the
Limit Of Quantication (LOQ), and the distribution of concentrations in the corresponding authentic model.


A.P. Minoja, C. Napoli / Food Research International 63 (2014) 126131

Of course, because the nutraceuticals world is so diverse, and the

related quality control problems are many and different, they require
targeted and/or non-targeted approaches for evaluation (for some
examples see Fujimura et al., 2011; Wang et al., 2005; Xiao, Dai,
Liu, Wang, & Tang, 2008). To give an idea on how NMR could help
in such eld, a couple of recent applications developed in collaboration with the Bruker application group are chosen and described.
Chlorogenic acid and hyperoside are found in signicant concentration in blueberry leaf extract and are regarded among the active
components responsible for the different healthy benets of such
herbal medicines (i.e. lowering blood sugar and triglyceride levels,
regulating glucose, and anti-inammatory). Consequently, it is important to have analytical methods for screening blueberry leaf extract
samples that identify and quantify these two components. This can be
done by HPLC, but it is quite laborious, time consuming and poorly
reproducible. Consequently, a single laboratory validation study
(Hicks et al., 2012) was carried out to develop an NMR methodology
that could alleviate the limits of the HPLC method. The NMR method
is based on the acquisition of a 1H spectrum and direct quantication,
by integration, of one peak for each component. Such a feasibility
study shows that the method is as precise, accurate and robust as the
HPLC approach yet it is simpler and faster. It must be taken into account
that sample handling is easy and the acquisition of one spectrum allows
for the quantication of both the investigated components simultaneously. The time of the data acquisition itself, depends on the signal
to noise ratio (S/N) required to reach a certain degree of precision and
accuracy, typically about 5%, but lowering to less than 1%. The achieved
S/N depends on the sample quantity, number of scans and the spectrometer conguration. Consequently, to make a realistic comparison,
spectra were acquired at 9 T (corresponding at 400 MHz proton resonance frequency, now considered an entry level magnetic eld
strength) and the experimental time was kept lower than one HLPC
run. The results of such study are so promising to deserve a deeper
investigation started with a multilaboratory trial to establish if quantitative analysis and multivariate methods could be applied across laboratories with different instrumentations (Markus et al., submitted for
Among the herbal treatments Ginseng, Panax spp., is one of the most
used (and consequently one the most sold!) because it is regarded as
effective against many ailments and for general well-being. The botanical
name Panax derives from the Greek (all) (heal). The investigation of the Ginseng bioactivity is rather complicated because it depends
on the type of Ginseng where it must be taken into account that there
are two main species of Ginseng, one native to North America, the
other one native to Asia. Additionally, local growing conditions affect
the product by altering the quantity of the various components like
polyacetylenes, triterpenes, polysaccharides, fatty acids, carbohydrates and vitamins (Assinewe, Baum, Gagnon, & Arnason, 2003).
Even if Ginseng bioactivity is related to either polysaccharide or to the
ginsenosides, it must be understood which molecules of these classes,
and how they are involved into the pharmacological process, in order
to prove the therapeutic benets.
Consequently, a NMR-based metabolomics approach (Yuk et al.,
2013) was evaluated to distinguish ve Ontario Ginseng landraces as
compared to each other and to discriminate between Ontario Ginseng
and Asian Ginseng. After the acquisition of 1H spectrum of each sample,
the average NMR spectrum was calculated for each landrace and for
Asian Ginseng. A visual comparison of the averaged spectra showed
peak variations in the metabolite prole, especially in the carbohydrate
region. Using a multivariate approach, Principal Component Analysis
(PCA) was additionally applied to discriminate the different materials
and to gain insight into the metabolic prole variations. The PCA distinguished clearly between the Asian and North American species showing
different maltose, sucrose and overall ginsenoside concentrations that
could be potential markers for discriminating the two species. Differentiation of the ve landraces, all of the North American species and

grown on farms in very close proximity, showed the ability of NMR to

show some distinction of these very closely related Ginseng plants.
In order to convert such methodology into a Ginseng quality control
tool further investigation is required, extending the number of samples
and developing classication models with minimum and maximum
condence intervals to recognize landrace and species of unknown
Ginseng samples or to verify correctness of the declared landrace and
4. Conclusions
NMR has proven to be a suitable technique for Fruit Juice and Wine
quality control. The same methodology is now ready to be transferred to
other food matrices. Non-targeted analysis is especially promising as it
brings a completely new food quality control approach by revealing unexpected responses rather than chasing already known problems.
Moreover, it should not be overlooked that targeted analysis allows
the determination of many parameters from a single experiment
resulting in a lower cost per parameter as compared to other analytical
technologies. An additional benet to NMR is its high reproducibility
that enables a single NMR instrument to be used to conduct many
different analyses without the need for additional calibration or adjustment for the specic material. This not only reduces the time of analysis
but also reduces the cost per sample.
As far as nutraceuticals quality control is concerned, the promise
shown by NMR for botanical screening, suggests that these highly reproducible and robust NMR methodologies are aptly suited to address
the urgent need for validated procedures for rapid quality assessment.
The described works have been developed by Bruker in collaboration with different research groups. The authors are grateful to all the
people involved, especially to:
P. Rinke and S. Koswig from SGF for the SGF methodology,
C. Kost, and F. Langenwalter from WineSpin Analytic; R. Godelmann,
H. Kuballa and Y. Monakhova from CVUA Karlsruhe; and N. Christoph
and H. Wachter from LGL Wrzburg for the wine methodology,
J. Ferrier and J.T. Arnason from University of Ottawa and Alain Currier
from Montreal Botanical Garden for the Blueberry study,
K.L. McIntyre, and J.T. Arnason from University of Ottawa; and D.
Brown, and E. Lui from Ontario Ginseng Innovation Research
Consurtium, for the Ginseng study.
The authors warmly thank colleagues Manfred Spraul and Kim
Colson for the continuous assistance and all the people from their
team, especially: H. Schaefer, B. Schutz, F. Fang, C. Cannet, D. Krings, A.
Steck M. Markus, J. Yuk, and S. Luchsinger.
Assinewe, V. A., Baum, B. R., Gagnon, D., & Arnason, J. T. (2003). Phytochemistry of wild populations of Panax quinquefolius L. (North American ginseng). Journal of Agricultural and
Food Chemistry, 51, 45494553.
Derome, A. E. (1987). Modern NMR techniques for chemistry research. Oxford: Pergamon
Fujimura, Y., Kurihara, K., Ida, M., Kosaka, R., Miura, D., Wariishi, H., et al. (2011).
Metabolomics-driven nutraceutical evaluation of diverse green tea cultivars. PLoS
ONE, 6, 116.
Hicks, J. M., Muhammad, A., Ferrier, J., Saleem, A., Cuerrier, A., Arnason, J. T., et al. (2012).
Quantication of chlorogenic acid and hyperoside directly from crude blueberry
(Vaccinium angustifolium) leaf extract by NMR spectroscopy analysis: Singlelaboratory validation. Journal of AOAC International, 95, 14061411.
Keeler, J. (2005). Understanding NMR spectroscopy. New York: John Wiley & Sons.
Lindon, J. C., Nicholson, J. K., Holmes, E., & Everett, J. R. (2000). Metabonomics: Metabolic
processes studied by NMR spectroscopy of biouids. Concepts in Magnetic Resonance,
12, 289320.
Mannina, L., Viel, S., & Sobolev, A. (2012). Liquid state 1H high eld NMR in food analysis.
Progress in Nuclear Magnetic Resonance Spectroscopy, 66, 139.
Markus, M. A., Luchsinger, S. M., Yuk, J., Ferrier, J., Cuerrier, A., Balick, M. J., et al. (2014).
Distinguishing Vaccinium species by chemical ngerprinting based on NMR spectra,

A.P. Minoja, C. Napoli / Food Research International 63 (2014) 126131

validated with spectra collected in different laboratories. Planta Medica (submitted
for publication).
Monakhova, Y. B., Kuballa, T., & Lachenmeier, D. W. (2012). Nontargeted NMR analysis to
rapidly detect hazardous substances in alcoholic beverages. Applied Magnetic
Resonance, 42, 343352.
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.
Nicholson, J. K., Lindon, J. C., & Holmes, E. (1999). Metabonomics: Understanding the
metabolic responses of living systems to pathophysiological stimuli via multivariate
statistical analysis of biological NMR data. Xenobiotica, 29(11), 11811189.
Saunders, J. K. M., & Hunter, B. K. (1993). Modern NMR spectroscopy. New York: Oxford
University Press.
Spraul, M., Schuetz, B., Humpfer, H., Moerterr, M., Schaefer, H., Koswig, S., et al. (2009). Mixture analysis by NMR as applied to fruit juice quality control. Magnetic Resonance in
Chemistry, 47, S130S137.


Wang, Y., tang, H., nicholson, J. K., hylands, P. J., Sampson, J., & Holmes, E. (2005). A
metabonomic strategy for the detection of the metabolic effects of chamomile
(Matricaria recutita L.) ingestion. Journal of Agricultural and Food Chemistry, 53,
Wider, G., & Dreier, L. (2006). Measuring protein concentrations by NMR spectroscopy.
Journal of the American Chemical Society, 128, 25712576.
Wrick, K. L. (2005). Impact of regulations on the business of nutraceuticals in the United States:
Yesterday, today and tomorrow. Regulation of functional foods and nutraceuticals: A
global perspective. Ames, Iowa: Blackwell Publishing, 1619.
Xiao, C., Dai, H., Liu, H., Wang, Y., & Tang, H. (2008). Revealing the metabonomic variation
of rosemary extracts using 1H NMR spectroscopy and multivariate data analysis.
Journal of Agricultural and Food Chemistry, 56, 1014210153.
Yuk, J., McIntyre, K. L., Fischer, C., Hicks, J., Colson, K. L., Lui, E., et al. (2013). Distinguishing
Ontario ginseng landraces and ginseng species using NMR-based metabolomics.
Analytical and Bioanalytical Chemistry, 405, 44994509.
Zeisel, S. H. (1999). Regulation of nutraceuticals. Science, 28, 18531855.