Ann. N.Y. Acad. Sci.

ISSN 0077-8923

A N N A L S O F T H E N E W Y O R K A C A D E M Y O F SC I E N C E S
Issue: Resveratrol and Health
CONCISE REVIEW

Resveratrol in the foodomics era: 1:25,000

Bekzod Khakimov and Søren Balling Engelsen
Chemometrics and Analytical Technology, Department of Food Science, University of Copenhagen, Frederiksberg, Denmark

Address for correspondence: Søren Balling Engelsen, Chemometrics and Analytical Technology, Department of Food
Science, University of Copenhagen, Rolighedsvej 26, 1958 Frederiksberg C, Denmark. se@food.ku.dk

Resveratrol is probably the most investigated plant secondary metabolite ever. An epidemiological study known as the
French paradox showed a correlation between red wine intake and low mortality due to coronary heart diseases, and
the red wine substance resveratrol was claimed to play a key role. Since then, several hundred resveratrol studies have
been conducted to demonstrate its antioxidant and other beneficial properties. In the foodomics era, considering a
complex foodome including over 25,000 substances that make up the human diet, it appears to be outdated to pursue
the hunt for biological activities one function/compound at a time. First, nature is multivariate, and the effect of any
one molecule will have to be modulated by its carrying matrix, its bioavailability, and synergies with other molecules.
Second, a large number of targeted studies have the tendency to become biased, as they tend to retain only the data
that the researchers think are relevant and thus increase the chances of spurious correlations. In this concise review,
we retrace the research toward a more inductive, holistic, and multivariate path.

Keywords: foodomics; metabolomics; chemometrics; resveratrol; inductive research

Introduction among other omics technologies, and there are still

a lot of food compounds and functions that remain
Resveratrol is one of approximately 25,000 com-
unexplored. Moreover, it is groundless to assume
pounds identified from foods to date.1–3 Food is
that a single molecule in a complex food matrix,
a complex system with drastically different physical
like red wine or grape juice, contributes to human
parameters, nutritional value, and chemical compo-
health and possesses the most efficient health bene-
sition. The foodome—a collection of all compounds
fits by acting alone.
(chemical substances) present in an investigated
Resveratrol is a plant secondary metabolite syn-
food sample at a given time—is immensely com-
thesized by stilbene synthase from the precursor
plex and dynamic, and relative concentrations of
molecule phenylalanine. Resveratrol, like other stil-
compounds determine properties of food, including
benes, is produced in plants in response to stressful
taste, smell, appearance, tenderness, and nutritional
conditions, including pathogens and UV radiation.
value. In fact, all foods are functional and, subse-
Grape skin is known to contain a particularly high
quently, chemical substances present in foods are
concentration of resveratrol, and red wine produced
also functional. The question is what compounds of
with a longer skin contact time possesses higher
foods are functional and possess health-beneficial
amounts of this phytoalexin.6 Resveratrol, which
effects. Resveratrol is a part of the foodome, and
has been known since ancient times in Chinese
despite the fact that it is one of the most studied
medicine as Hu Zhang, suddenly drew considerable
food metabolites, it is not necessarily the only one
public attention in the early 1990s after the publi-
and/or the most efficient metabolite for improving
cation of the epidemiological study known as the
human health and well-being. Foodomics—a disci-
“French paradox.”7 Renaud and Lorgeril reported
pline that studies the food and nutrition domains
that, despite high consumption of saturated fat and
through the application of advanced omics tech-
high blood cholesterol levels, which lead to coronary
nologies to improve the consumer’s well-being,
heart disease (CHD), the mortality from this disease
health, and knowledge4,5 —is a relatively young field
was notably lower among the French population
doi: 10.1111/nyas.13425
48 Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Khakimov & Engelsen
compared with the U.S. and UK populations with
similar fat intake and blood cholesterol levels. The
paradox was explained by high negative correla-
tion between wine intake and mortality from CHD,
where alcohol was associated with a reduction in
CHD cases. Later, this phenomenon was related to
the antioxidant activity of the red wine substrate
resveratrol, and it was argued that resveratrol can
reduce the human platelet aggregation that causes
CHD. The correlations illustrated by Renaud and
Lorgeril were all valid, but they were only correla-
tions, not causal relationships. Nevertheless, the cor-
relations raised many questions regarding whether
wine intake indeed reduces mortality from CHD,
since similar or even better correlation can easily be
achieved using variables other than wine intake (e.g.,
dietary and lifestyle habits as well as ecosystem).
The assumption made in the French paradox
paper attracted considerable public attention, and
several studies conducted after the discovery sug-
gested a role for resveratrol in preventing CHD.8–10
Resveratrol is probably the most-explored plant sec-
ondary metabolite ever. Many studies conducted on
resveratrol have covered a broad range of biomed-
ical sciences, and it was applied in several hundreds
of in vitro studies in order to investigate its phar-
macological properties, including antioxidant,11,12
anti-inflammatory,12 antimicrobial,13 antiviral,14
antiaging,15 antidiabetic,16 cancer chemopreventive
activity,48 and immunomodulatory activities17
(Fig. 1). In addition, animal model studies showed
that resveratrol has a high potential to be applied
as medicine against CHD and various cancers and
as an antiviral agent.18 The majority of these in
vitro and animal studies, mostly involving rodents,
however, were performed using abnormally large
concentrations of resveratrol that significantly
change the physiological behavior of exposed cells,
and thus evaluation of the true biological effect of
the compound remains elusive. Moreover, most
intervention studies were conducted for a short
time period and included no validation and/or
extrapolation of the observed results to humans. In
vitro and animal studies have shown evidences of
resveratrol for lowering bad low-density lipoprotein
(LDL) cholesterol,19,20 lowering blood glucose,
inhibiting cancer cell growth,21 and improving the
immune system. Resveratrol has therefore been
intensively studied in clinical trials targeted toward
cardiovascular22 and neurological diseases, as well as
Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Resveratrol in the foodomics era
toward cancer,23 HIV,24 diabetes,25 and obesity,26,27
but with no breakthrough application so far.
Human studies dealing with the pharmacoki-
netics and metabolism of resveratrol have so far
been limited to small cohorts, often including
fewer than 50 individuals.18,28,29 The majority of
the studies have been conducted to test its effect
in single cell lines and in animal models, normally
with a deductive focus on predefined hypotheses
and/or findings.30,31 By comparison, the number
of studies performed to understand the role of
resveratrol in human health has been very limited.
Thus, there is a big gap in arguments that must be
filled before the scientific community can prove the
actual role and biochemical mechanism of action
of resveratrol in humans. Some of the obvious
questions that remain open are (1) effect of resver-
atrol in human metabolism; (2) the bioavailability
and effects of resveratrol depending on gender,
age, health, lifestyle, diet, and ethnicity; (3) the
source of resveratrol for humans for its maximum
bioavailability; and (4) what resveratrol can be used
safely and effectively. It is outside the scope of this
paper to review and discuss in detail all studies
performed to date on humans involving resveratrol
and/or resveratrol-containing foods. Instead, we
summarize the most striking findings/claims about
the health effects of resveratrol in humans and,
most importantly, we propose how to set up an
exploratory human trial to gain unbiased knowl-
edge about resveratrol’s actions(s) in the human
body. The latter is demonstrated using an example
from a dietary intervention study performed
among a Danish population. We highlight the most
important aspects of human intervention studies
to be considered when performing comprehensive
metabolomics analysis. In addition, we discuss fac-
tors to be included in an intervention study design in
order to avoid confounding effect that may hamper
the ability to detect the effects sought. Advantages
and limitations of the state-of-the art multivariate
data analysis techniques will also be discussed in
order to illustrate how information can effectively
be extracted from the complex metabolomics data.
Current status of resveratrol research
The scientific interest in the so-called French para-
dox generated a large number of studies in vari-
ous areas of the biochemical research (Fig. 2). At
present, more than 10,000 scientific publications
49
Resveratrol in the foodomics era Khakimov & Engelsen

Figure 1. An overview of pharmaceutical properties (antioxidant,1,2 antimicrobial,3,4 anti-inflammatory,4 antiviral7,8 ), biological

effects (lowers blood pressure,13,14 lowers LDL and triglycerides,13,14 lowers blood sugar,15,16 boosts immune system17,18 ), and
clinical investigations (cardiovascular diseases,9,10 cancer,11,12 diabetes,13,14 HIV,15,16 obesity17,18 ) of resveratrol.

can be found in the literature using the key word
“resveratrol.” It has been shown that this single com-
pound possesses multiple biological activities and
that it is able to interact with cells once it makes
first contact. It is also indisputable that resvera-
trol can scavenge free radicals and prevent unde-
sirable oxidation reactions in cells.11,12 Owing to its
unique chemical structure, it can easily pass through
cell membranes and interact with enzymes and/or
inhibit growth and development of microorgan-
isms and viruses.14 Despite the fact that majority
of these findings have used in vitro and animal
model approaches, one cannot reject the hypothesis
that this single molecule can be used for improv-
ing human health or be applied to cure a certain
disease. However, at present, such an application
of resveratrol is impossible owing to many reasons.
First, the scientific community still cannot eluci-
date a single exact mode of action of resveratrol in
the human body and how it perturbs metabolism.
However, one can question whether it is a necessity
to know the exact mechanism of resveratrol if its

50 Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 

C 2017 New York Academy of Sciences.
Khakimov & Engelsen
Figure 2. The “resveratrol paradox” pyramid.
health-promoting effects have already been proven.
Nevertheless, understanding how it works in the
human body may reveal new knowledge, broaden its
applications, and avoid possible side effects. Second,
we still do not have enough data about its bioavail-
ability and what factors dictate it, since it was shown
that the bioavailability of resveratrol largely varies
between persons and food sources (e.g., the bioavail-
ability of resveratrol in red wine is higher than
when it is administered in a tablet form).18 Third,
it is unclear whether resveratrol or its glycosylated
derivatives are the functional form. More impor-
tantly, there are no convincing studies showing their
safety for human health or showing any poten-
tial side effects. Last but not least, what is the dis-
ease(s) or symptom(s) that could be cured using safe
amounts of resveratrol in humans? In this respect,
it should be mentioned that most resveratrol inter-
ventions have been conducted using extraordinary
high concentrations that largely exceed the typical
amounts that one may consume by drinking red
wine or eating grapes.
Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Resveratrol in the foodomics era
Most of above-mentioned questions are partially
addressed in various scientific papers dealing
with human trials,18,32,33 where, unfortunately,
the findings are sometimes contradictory. Health
effects, bioavailability data, and dose-dependent
response of resveratrol greatly vary in studies,
and this may indicate individualized response
to resveratrol. However, it is also possible that
these contradictory findings are the result of
inappropriate study designs. Several resveratrol
human trials have been designed to measure only
a few variables, and the possible conclusions from
such restricted data are limited. In the foodomics
age, such targeted and limited analyses contradict
the inherent complexity of the biological systems,
which require simultaneous measurement of as
many parameters as possible and subsequent
multivariate data analysis in order to extract the
maximum information. We propose that the
following six points be considered in future human
studies for investigation of the health effects of
resveratrol.
51
Resveratrol in the foodomics era Khakimov & Engelsen

One size does not fit all dinal matrix (i.e., red wine, grape juice, and other
The phenotypic diversity of humans requires that foods). In this respect, the hydrophilicity is a key
a larger number of subjects must be recruited for factor: is it water soluble or fat soluble? The glycosy-
a general-purpose intervention study and that the lated derivatives of resveratrol are more hydrophilic
study takes into account parameters such as gender, and bulkier than resveratrol itself and may have
age, health status, diet, lifestyle, and ethnicity.34 very different bioavailability and biological activi-
ties, which has been only scarcely studied to date.
Experimental design
Second, if the target is to emulate epidemiological
The experimental design is often the limiting fac-
studies, the supplement dose should be close to the
tor in most biological experiments. It is important
natural concentrations to which people have been
to use an appropriate study design, preferably bal-
exposed. Abnormally high doses can be used to pro-
anced, such that equal numbers of subjects are inves-
voke a reaction, but the interpretation will be more
tigated for each level of each study factor. Human
difficult.
intervention studies normally use a power calcula-
tion to determine the number of subjects, the daily
The health definition
dose employed as supplement, the duration of the
According to the World Health Organization
intervention, and the number of technical replicates
(WHO), health is defined as a “state of complete
that need to be included at each design level in order
physical, mental, and social well-being, and not
to obtain a minimum statistically significant effect.
merely the absence of disease or infirmity.” Health
This calculation includes the expected effect and
is a dynamic condition resulting from a body’s
the magnitude of the analytical variances. While
constant adjustment and adaptation in response
this calculation is more or less straightforward in
to stresses and changes in the environment for
univariate experiments, it often falls short for mul-
maintaining an inner equilibrium called homeosta-
tivariate and multimetabolite experiments, such as
sis. Accordingly, any health-promoting substance
is the standard in nutritional metabolomics. This
should be able to improve our response to any given
is why the size and design of many nutrition inter-
food challenge (e.g., how quickly the body is able to
vention studies have been decided by gut feeling,
reach homeostasis).
but obviously a more reliable and efficient way of
estimating sample size in large intervention studies
An inductive measurement strategy
is to perform an appropriately designed pilot study
It has been a scientific convention to measure only
before conducting the large experiment.34 Such a
one clinical variable of interest as the response to
pilot study involving fewer subjects can help deter-
one compound of interest (e.g., resveratrol) in a so-
mine expected variances corresponding to the study
called deductive measurement strategy. However,
design factors as well as variances derived from
if you only measure what you think is relevant,
experimental error using technical replicates. Sub-
then you will only confirm what you already know
sequently, these variances can be used to calculate
and may lose a chance of discovering unforeseen
minimum statistically significant effects and sam-
effects and hinder innovation. New developments
ple sizes for the large intervention study. Very often,
in analytical technology and multivariate data anal-
human intervention studies designed to investigate
ysis facilitate the use of an inductive measurement
the effects of a certain diet or food ingredient(s)
strategy in which advanced analytical equipment is
will employ hundreds to a few thousand subjects.
used to “shake the tree” and advanced multivariate
Moreover, most modern metabolomics studies use
data-mining methods are used to find underlying
kinetic sampling with multiple measurement time
metabolite patterns that may explain the complex
points rather than static end-point measurement,
biology behind the scene. The uniqueness of the
which enhances the mechanistic interpretation.
inductive strategy will be demonstrated by a case
Matrix effects study (vide supra) in which an intervention with
The biological effects of any food compound depend a Nordic diet was compared to an average Danish
on its bioavailability: does it reach its target? First, it diet (ADD) and a surprising result was discovered,
is thus important to complement any investigation which by no means could have been discovered in a
of pure resveratrol with resveratrol in its habitu- deductive approach.

52 Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 

C 2017 New York Academy of Sciences.
Khakimov & Engelsen
Nature is multivariate
It has been demonstrated again and again that
nature is multivariate, and thus new experimen-
tal designs, including many variables, should be
explored through more holistic multivariate meth-
ods rather than being limited by univariate anal-
ysis methods. As an example, an organism can-
not survive by only having one of the essential
amino acids in its feed; it will need a complex
mixture (pattern) of all the essential amino acids.
When it comes to nonessential but beneficial com-
pounds, it is most likely that molecular networks
and molecular affinities will create even more com-
plex underlying patterns—also called the cage of
covariation. The chance for the experimental design
to result in a pattern of compounds that classify
the effects of the intervention is way beyond the
1:25,000 chance for resveratrol. To this end, power-
ful multivariate data-mining technologies have been
developed, including principal component analysis
(PCA),35 partial least squares–discriminant analy-
sis (PLS-DA),36 and analysis of variance (ANOVA)
simultaneous component analysis (ASCA).37
The last two points are indeed the least applied in
current resveratrol research, and, surprisingly, most
studies are deductive and demonstrate predefined
hypotheses (e.g., resveratrol suppresses growth of
cancer cell lines by 25%). In contrast, questions like
“What has happened to the cell’s metabolism?” and
“What other molecules in cells are affected by resver-
atrol?” remain scarce. Nevertheless, a small number
of studies have attempted to apply top-down holis-
tic metabolomics approaches. However, nearly all
metabolomics studies that have been performed to
evaluate the effects of resveratrol were conducted in
in vitro systems using abnormally high concentra-
tions of the compound and often targeted a minor
part of the cells’ metabolome.18,38 As an example,
one of the few metabolomics studies, using liquid
chromatography mass spectrometry (LC-MS) con-
ducted in human breast cancer cell lines, showed
that resveratrol inhibits the growth of cancer cells
and significantly alters cell metabolism, which was
more pronounced in amino acid synthesis.39 In this
case, the applied doses of resveratrol were very high,
which makes it challenging to estimate how cells
would behave if they received a lower (safe for
human consumption) concentration, and if in real
life resveratrol can inhibit human breast cancer cell
growth in the same way. Likewise, a nuclear mag-
Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Resveratrol in the foodomics era
netic resonance (NMR) metabolomics study per-
formed on human liver cancer cell lines (HepG2)
showed an effect of resveratrol in glucose and amino
acid utilization to fat utilization for the production
of energy and lowered the rate of the cancer cell
cycle.40 One recent study investigating metabolic
effects of resveratrol was performed on men with
metabolic syndrome and included a 4-month inter-
vention followed by comprehensive metabolome
analyses in four sample matrices—blood, urine, fat,
and muscle by—LC-MS and gas chromatography
(GC)-MS platforms.33 This study showed that lipid
metabolism in muscle was affected, which resulted
in increased intracellular glycerol and accumula-
tion of long-chain free fatty acids in response to
resveratrol, while the urine metabolome exhibited
increased levels of aromatic amino acids. Despite the
fact that this work is one of the most comprehen-
sive metabolomics studies of resveratrol treatments,
the data generated have not been fully explored
and limited to univariate analysis. Another impor-
tant point is the bioavailability and metabolism of
resveratrol in humans, which raises several ques-
tions regarding its health effects. The human inter-
vention study where healthy individuals received red
wine containing 0.8 or 3.2 ␮g/mL trans-resveratrol
showed that only four of 25 individuals exhibited
trace amounts of free resveratrol in their serum sam-
ples 30 min after red wine intake,41 while only nine
of 25 individuals possessed 3- and 4-glucuronide
trans-resveratrol 0.5–2 h after red wine intake. How-
ever, considering the complexity of the design (e.g.,
gender and body mass difference, consumed resver-
atrol concentration, and diet), this study included
a rather low number of participants, which chal-
lenges the ability to evaluate the effects, as the data
may be dominated by interindividual variations
in metabolic systems and in resveratrol bioavail-
ability. Therefore, the scientific community dealing
with bioactive compounds is now urging that more
research be conducted to demonstrate the bioavail-
ability of bioactive constituents in relevant in vitro
and in vivo models.42,43
How to design an ideal resveratrol
intervention study
As has been shown in previous studies, the bioavail-
ability of resveratrol largely depends on whether it
is consumed via drinking red wine, liquor, or juice,
and its further metabolism displays even greater
53
Resveratrol in the foodomics era
Figure 3. A typical workflow of a metabolomics study conducted to investigate the effects of a single medicine (e.g., resveratrol),
food supplement, probiotic, single food product, or entire diet on recovery rate of a certain disease, human metabolism, health,
well-being, and life expectancy. 
C Engelsen & Newlin.
variation among participants.41,44 Thus, one should
consider taking into account the multivariate nature
of the foodome and strong relationships present
among compounds, since it is impossible to iso-
late their digestion from all other interactions in the
human body. In order to validate the health claims
made about resveratrol and to extrapolate the bio-
logical effects found in in vitro and animal studies to
humans, one must not only measure resveratrol and
its metabolites, but also conduct untargeted analy-
sis. Although such experiments are more expen-
sive and laborious and require expert knowledge in
multidisciplinary areas, such as metabolomics and
multivariate data analysis, it is the only viable way
to see the holistic picture (Fig. 3). It is difficult or
even impossible to fully understand what happens
to resveratrol once it reaches our blood and how it
is further metabolized and interacts with other cell
components. This challenge becomes more realistic
when untargeted metabolomics data are combined
with all other clinical and metadata generated (e.g.,
in a resveratrol intervention study). Such a fusion
of data may reveal unforeseen correlations that in
fact carry new biological information and augment
the current knowledge of resveratrol’s role in human
metabolism and health. More importantly, compre-
hensive foodomics studies, including genome-wide
association studies,4 have even higher chances to
obtain novel biological insight by associating the
genetic make-up with the metabolic effects of bioac-
tive compounds, which can point to personalized or
stratified diets. However, the current status of the
54
Khakimov & Engelsen
technology as well as research interests are mostly
limited to evaluating only one type of omics vari-
able, which prevents linking the knowledge com-
ing from the gene level to all the way down to the
metabolite and functional levels.
The state-of-the-art analytical technologies,
including GC–MS, LC–MS, and NMR, applied in
foodomics and metabolomics45,46 enable simult-
aneous detection of up to a few hundred metabo-
lites from a single analysis, which reflects the
current physiological status of the investigated
biological samples. For example, the blood plasma
metabolome measured by GC–MS at the fasting
state provides relative or absolute concentrations of
up to 200 metabolites in a single run, which provides
a metabolic fingerprint of the subject in home-
ostasis. If that subject then undergoes resveratrol
treatment, and it is possible to collect the subject’s
blood with a frequency that allows for monitoring
of resveratrol metabolism, it will be possible to
evaluate which compounds change owing to the
resveratrol intake and when and by how much their
relative concentrations change. Similarly, these
blood plasma GC–MS data can easily be reflected the
health parameters (e.g., blood pressure or glucose
or insulin levels). Such a top-down approach may
lead to the discovery of new biochemical processes
and allow evaluation of fast metabolic changes that
are either directly or indirectly related to each other.
In order to perform comprehensive and untar-
geted metabolomics studies, one must carefully
design the experiment to allow enough power for
Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Khakimov & Engelsen Resveratrol in the foodomics era

Figure 4. Decomposition of variations in hypothetical data, with I subjects and J variables, derived from a balanced resveratrol
study using ANOVA-simultaneous component analysis (ASCA). This study illustrates a fully balanced design including three main
effects: gender (at two levels), treatment (at three levels), and time (at three levels). ASCA allows evaluation of effect size for each
main effect individually as well as their two- and three-factor interaction effects.

estimation of the effects and/or variations expected.
Thus, it is crucial to consider all six points men-
tioned above. It is worth mentioning that many
studies in the literature explore generated data in a
very limited way and often do not extract maximum
information. This is mainly due to inappropriately
designed studies that lead to confounding of effects
and/or lack of expertise in multivariate data anal-
ysis. However, properly designed human interven-
tions with sufficient numbers of subjects per study
group and carefully conducted experiments allow
accurate estimation of the effects of all primary fac-
tors as well as their two-factor and higher order
interaction effects. One such approach is ASCA.
ASCA is a state-of-the-art method for partitioning
sources of variation in the data derived from study
design factors and their interactions. ASCA parti-
tions variations derived from different factors in the
same way as classical ANOVA, but accounts for all
variables included in a data simultaneously. Let us
assume a human intervention study with resver-
atrol where three main effects—gender, treatment,
and time—are included in the study design, and this
investigation results in a data matrix X(I × J) with I
samples and J variables (Fig. 4). Assuming that the
data matrix X is balanced—there are equal numbers
of subjects in each level for all three factors—one
can write an ANOVA equation that will account
for seven effect matrices, three main effects, three
two-factor interaction effects, and one three-factor
interaction effect as
X = X MEAN + X GENDER + X TREATMENT + X TIME
+ X GENDER×TREATMENT
+ X GENDER×TIME + X TREATMENT×TIME
+ X GENDER×TREATMENT×TIME + X RESIDUALS (1)
where XMEAN is a matrix of means (the same size
as X) that is generated from the overall means of
all variables in the data. For a balanced design,
the variance partitioning reveals effect matrices
XGENDER , XTREATMENT , XTIME , XGENDER × TREATMENT ,
XGENDER × TIME , XTREATMENT × TIME , and
XGENDER × TREATMENT × TIME that are orthogo-
nal to each other. During ASCA, all of these
effect matrices will be generated, and each of
them can be validated using a permutation test by

Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 

C 2017 New York Academy of Sciences. 55
Resveratrol in the foodomics era Khakimov & Engelsen

calculating sum of squares of the effect matrices. application of PLS-DA-based variable selection.36
This way of decomposing the variations to the One of the unforeseen metabolite markers for the
design factors and their interaction terms not diet was 2,6-diisopropylnaththalene, an antisprout-
only allows estimation of each effect size but also ing agent used for long-term storage of potatoes
determination of whether these effects are real for the production of chips, which was significantly
(statistically significant). For a balanced study elevated in the blood of ADD individuals.
design, these variances are unambiguous. Each
effect matrix can be investigated separately by Outreach
performing a local PCA analysis,35 where the
In many ways, we are living in a very fragmented
nondesign-related variation (X RESIDUALS in Eq.
world, hunting for causality that for the moment
(1)) is used as a measure of uncertainty. However,
can provide answers. This worldview leads to side
if the generated data are unbalanced (e.g., there are
effects because of neglected covariance and context,
more females in the intervention than males), the
as is often seen causing, for example, environmental
main effects remain valid while interactions are no
problems. In the case of resveratrol, the neglected
longer valid. In order to cope with this, one may
covariance and context related to red wine con-
consider reducing the data size to make it balanced
sumption have either led us to the wrong conclusion
and/or utilize an iterative approach by which a
or caused us to miss one or more important pieces
few sets of samples from the exceeding blocks of
of the French paradox puzzle. Foodomics can be
the study design are removed in a random order
considered as a door opening research for future
to generate balanced data from unbalanced data,
studies on the development and improvement of
which are then evaluated by ASCA; this procedure
healthy food products and ingredients according to
can be repeated n times for n balanced data sets to
the European Food Safety Authority’s regulations.
estimate interaction.47 This approach for dealing
Indeed, inductive foodomics studies may facilitate
with unbalanced study designs allows validation of
generation of more specific goals and identify highly
interaction terms without permanently reducing
promising and potential bioactive foods and/or sin-
the sample size.
gle bioactive substances. Despite the large number
In contrast to common unsupervised multivari-
of studies that have been conducted on the possi-
ate data analysis methods, such as PCA, which finds
ble health and other biological effects of resvera-
common underlying variations in the data, ASCA
trol, there is still a need for more systematic and
provides decomposition of variations on the basis
detailed investigations to understand resveratrol’s
of the study design. The latter allows identification
role in human metabolism and health. In order to
of the small fraction of the biological variation that is
avoid future wastes of research efforts, we urge the
otherwise hidden under more abundant individual
resveratrol research community to reject deductive
variation in large human cohorts. One such exam-
approaches and move toward more holistic induc-
ple is the dietary intervention study conducted to
tive studies. The latter can be successfully performed
evaluate human blood plasma metabolic responses
by applying modern analytical advances developed
to long-term effect of the New Nordic diet (NND)
in the fields of metabolomics and foodomics. Ulti-
versus the ADD.47 In this study application, ASCA
mately, this multidisciplinary foodomics approach
allowed partitioning of the variations derived from
may reveal a holistic overview of the effects of resver-
factors of the study design, diet, season, and gen-
atrol in human metabolism and elucidate its mode
der and enabled separation of the diet effect, which
of action, which in turn will result in development
accounted for only 1% of the total variation present
of targeted applications of resveratrol in the phar-
in the metabolomics data. Although the diet effect
maceutical and food industries.
was small, it remained significant, while most of
the variation (>90%) was kept in the residuals,
Acknowledgments
which comprised individual variations as well as
other variations not related to the study design. This The authors wish to acknowledge the Dan-
approach allowed unbiased evaluation of whole diet ish Innovation Foundation for generous sup-
and resulted in metabolite markers for diet, season, port to the project entitled “COUNTERSTRIKE—
and gender differences that were validated by the Counteracting Sarcopenia with proteins and

56 Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 

C 2017 New York Academy of Sciences.
Khakimov & Engelsen
exercise—screening the CALM cohort for lipopro-
tein biomarkers” (IFD: 4105-00015B).
Competing interests
The authors declare no competing interests.
References
1. Scalbert, A. et al. 2014. The food metabolome: a window
over dietary exposure. Am. J. Clin. Nutr. 99: 1286–1308.
2. Khakimov, B., B.M. Jespersen & S.B. Engelsen. 2014.
Comprehensive and comparative metabolomic profiling of
wheat, barley, oat and rye using gas chromatography–mass
spectrometry and advanced chemometrics. Foods 3: 569–
585.
3. Khakimov, B. et al. 2016. A comprehensive and comparative
GC–MS metabolomics study of non-volatiles in Tanzanian
grown mango, pineapple, jackfruit, baobab and tamarind
fruits. Food Chem. 213: 691–699.
4. Cifuentes, A. 2012. Food analysis in the postgenomic era:
foodomics. Electrophoresis 33: 2199–2200.
5. Khakimov, B., G. Gürdeniz & S.B. Engelsen. 2015. Trends in
the application of chemometrics to foodomics studies. Acta
Aliment. 44: 4–31.
6. Yama, U.R. et al. 2016. Effect of vegetation time and climatic
conditions on trans-resveratrol concentrations in Caber-
net Sauvignon and Merlot Wines from different regions in
Turkey. S. Afr. J. Enol. Vitic. 37: 85–92.
7. Renaud, S. & M. Delorgeril. 1992. Wine, alcohol, platelets,
and the French paradox for coronary heart disease. Lancet
339: 1523–1526.
8. Frankel, E.N., A.L. Waterhouse & J.E. Kinsella. 1993. Inhi-
bition of human LDL oxidation by resveratrol. Lancet 341:
1103–1104.
9. Paceasciak, C.R. et al. 1995. The red wine phenolics trans-
resveratrol and quercetin block human platelet-aggregation
and eicosanoid synthesis: implications for protection against
coronary heart disease. Clin. Chim. Acta 235: 207–219.
10. Wu, J.M. et al. 2001. Mechanism of cardioprotection by
resveratrol, a phenolic antioxidant present in red wine
(Review). Int. J. Mol. Med. 8: 3–17.
11. Santos, J.A. et al. 2013. Resveratrol and analogues: a review
of antioxidant activity and applications to human health.
Recent Pat. Food Nutr. Agric. 5: 144–153.
12. Lancon, A., R. Frazzi & N. Latruffe. 2016. Anti-oxidant, anti-
inflammatory and anti-angiogenic properties of resveratrol
in ocular diseases. Molecules 21: 304.
13. Maeurer, M., M. Rao & A. Zumla. 2016. Host-directed ther-
apies for antimicrobial resistant respiratory tract infections.
Curr. Opin. Pulm. Med. 22: 203–211.
14. Yang, T. et al. 2015. Resveratrol, sirtuins, and viruses. Rev.
Med. Virol. 25: 431–445.
15. Baxter, R.A. 2008. Anti-aging properties of resveratrol:
review and report of a potent new antioxidant skin care
formulation. J. Cosmet. Dermatol. 7: 2–7.
16. Szkudelski, T. & K. Szkudelska. 2015. Resveratrol and dia-
betes: from animal to human studies. Biochim. Biophys. Acta
1852: 1145–1154.
Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
C 2017 New York Academy of Sciences.
Resveratrol in the foodomics era
17. Jantan, I., W. Ahmad & S.N.A. Bukhari. 2015. Plant-derived
immunomodulators: an insight on their preclinical evalua-
tion and clinical trials. Front. Plant Sci. 6: 655.
18. Tome-Carneiro, J. et al. 2013. Resveratrol and clinical trials:
the crossroad from in vitro studies to human evidence. Curr.
Pharm. Des. 19: 6064–6093.
19. Voloshyna, I., S.M. Hussaini & A.B. Reiss. 2012. Resveratrol
in cholesterol metabolism and atherosclerosis. J. Med. Food
15: 763–773.
20. Bradamante, S., L. Barenghi & A. Villa. 2004. Cardiovascular
protective effects of resveratrol. Cardiovasc. Drug Rev. 22:
169–188.
21. Singh, C.K., M.A. Ndiaye & N. Ahmad. 2015. Resveratrol and
cancer: challenges for clinical translation. Biochim. Biophys.
Acta 1852: 1178–1185.
22. Petrovski, G., N. Gurusamy & D.K. Das. 2011. Resveratrol in
cardiovascular health and disease. Ann. N.Y. Acad. Sci. 1215:
22–33.
23. Alayev, A., S.M. Berger & M.K. Holz. 2015. Resveratrol as a
novel treatment for diseases with mTOR pathway hyperac-
tivation. Ann. N.Y. Acad. Sci. 1348: 116–123.
24. Clouser, C.L. et al. 2012. Anti-HIV-1 activity of resveratrol
derivatives and synergistic inhibition of HIV-1 by the com-
bination of resveratrol and decitabine. Bioorg. Med. Chem.
Lett. 22: 6642–6646.
25. Hausenblas, H.A., J.A. Schoulda & J.M. Smoliga. 2015.
Resveratrol treatment as an adjunct to pharmacological
management in type 2 diabetes mellitus-systematic review
and meta-analysis. Mol. Nutr. Food Res. 59: 147–159.
26. Aguirre, L. et al. 2014. Resveratrol: anti-obesity mechanisms
of action. Molecules 19: 18632–18655.
27. de Ligt, M., S. Timmers & P. Schrauwen. 2015. Resveratrol
and obesity: can resveratrol relieve metabolic disturbances?
Biochim. Biophys. Acta 1852: 1137–1144.
28. Bedada, S.K., N.R. Yellu & P. Neerati. 2016. Effect of resver-
atrol treatment on the pharmacokinetics of diclofenac in
healthy human volunteers. Phytother. Res. 30: 397–401.
29. Brown, V.A. et al. 2010. Repeat dose study of the can-
cer chemopreventive agent resveratrol in healthy volun-
teers: safety, pharmacokinetics, and effect on the insulin-like
growth factor axis. Cancer Res. 70: 9003–9011.
30. Yang, Z.Q. et al. 2016. Preclinical pharmacokinetics
comparison between resveratrol 2-hydroxypropyl-beta-
cyclodextrin complex and resveratrol suspension after oral
administration. J. Incl. Phenom. Macrocycl. Chem. 86: 263–
271.
31. Kapetanovic, I.M. et al. 2011. Pharmacokinetics, oral
bioavailability, and metabolic profile of resveratrol and
its dimethylether analog, pterostilbene, in rats. Cancer
Chemother. Pharmacol. 68: 593–601.
32. Erdogan, C.S. & O. Vang. 2016. Challenges in analyzing the
biological effects of resveratrol. Nutrients 8. pii: E353
33. Korsholm, A. et al. 2017. Comprehensive metabolomic anal-
ysis in blood, urine, fat, and muscle in men with metabolic
syndrome: a randomized, placebo-controlled clinical trial
on the effects of resveratrol after four months’ treatment.
Int. J. Mol. Sci. 18: 554.
34. Trutschel, D. et al. 2015. Experiment design beyond gut feel-
ing: statistical tests and power to detect differential metabo-
lites in mass spectrometry data. Metabolomics 11: 851–860.
57
Resveratrol in the foodomics era
35. Hotelling, H. 1933. Analysis of a complex of statistical vari-
ables into principal components. J. Educ. Psychol. 24: 417–
441.
36. Ståhle, L. & S. Wold. 1987. Partial least squares analysis with
cross-validation for the two-class problem: a Monte Carlo
study. J. Chemometrics 1: 185–196.
37. Smilde, A.K. et al. 2005. ANOVA-simultaneous compo-
nent analysis (ASCA): a new tool for analyzing designed
metabolomics data. Bioinformatics 21: 3043–3048.
38. Rotches-Ribalta, M. et al. 2012. Gut and microbial resver-
atrol metabolite profiling after moderate long-term con-
sumption of red wine versus dealcoholized red wine in
humans by an optimized ultra-high-pressure liquid chro-
matography tandem mass spectrometry method. J. Chro-
matogr. A 1265: 105–113.
39. Jager, W. et al. 2011. Metabolomic analysis of resveratrol-
induced effects in the human breast cancer cell lines MCF-7
and MDA-MB-231. OMICS 15: 9–14.
40. Massimi, M. et al. 2012. Effects of resveratrol on HepG2
cells as revealed by (1)H-NMR based metabolic profiling.
Biochim. Biophys. Acta 1820: 1–8.
41. Vitaglione, P. et al. 2005. Bioavailability of trans-resveratrol
from red wine in humans. Mol. Nutr. Food Res. 49:
495–504.
58 Ann. N.Y. Acad. Sci. 1403 (2017) 48–58 
Khakimov & Engelsen
42. Somoza, V. et al. 2015. Guidelines for research on bioactive
constituents—a journal of agricultural and food chemistry
perspective. J. Agric. Food Chem. 63: 8103–8105.
43. Gonzales, G.B. 2016. In vitro bioavailability and cellular
bioactivity studies of flavonoids and flavonoid-rich plant
extracts: questions, considerations and future perspectives.
Proc. Nutr. Soc. 1–7.
44. Walle, T. 2011. Bioavailability of resveratrol. Ann. N.Y. Acad.
Sci. 1215: 9–15.
45. Khakimov, B., S. Bak & S.B. Engelsen. 2014. High-
throughput cereal metabolomics: current analytical tech-
nologies, challenges and perspectives. J. Cereal Sci. 59: 393–
418.
46. Sørensen, K.M., B. Khakimov & S.B. Engelsen. 2016. The use
of rapid spectroscopic screening methods to detect adulter-
ation of food raw materials and ingredients. Curr. Opin. Food
Sci. 10: 45–51.
47. Khakimov, B. et al. 2016. New Nordic diet versus average
Danish diet: a randomized controlled trial revealed healthy
long-term effects of the new Nordic diet by GC–MS blood
plasma metabolomics. J. Proteome Res. 15: 1939–1954.
48. Jang, M., et al. 1997. Cancer chemopreventive activity of
resveratrol, a natural product derived from grapes. Science
275: 218–220.
C 2017 New York Academy of Sciences.