Cell Metabolism

Perspective

Systems Biology of Metabolism: A Driver

for Developing Personalized and Precision Medicine
Jens Nielsen1,2,3,*
1Department of Biology and Biological Engineering, Chalmers University of Technology, SE41128 Gothenburg, Sweden
2Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, DK2800 Lyngby, Denmark
3Science for Life Laboratory, Royal Institute of Technology, SE17121 Stockholm, Sweden

*Correspondence: [email protected]
http://dx.doi.org/10.1016/j.cmet.2017.02.002

Systems biology uses mathematical models to analyze large datasets and simulate system behavior. It
enables integrative analysis of different types of data and can thereby provide new insight into complex
biological systems. Here will be discussed the challenges of using systems medicine for advancing the
development of personalized and precision medicine to treat metabolic diseases like insulin resistance,
obesity, NAFLD, NASH, and cancer. It will be illustrated how the concept of genome-scale metabolic models
can be used for integrative analysis of big data with the objective of identifying novel biomarkers that are
foundational for personalized and precision medicine.

Introduction participatory. The concept of preventive, predictive, personal-

Healthcare costs are rapidly increasing in the developing coun- ized, and participatory medicine has been coined P4 medicine
tries, and in 2011 the total healthcare spending in the United (Hood and Friend, 2011), and this may completely transform
States accounted for about 18% of its GDP (WHO, 2011), a the healthcare sector in the next 1020 years.
63% inflation-adjusted increase since 1997 (Pfuntner et al., Through N-of-1 trials on a large number of people, it will
2011). Despite this, many people are taking drugs that will not become possible to develop more detailed data-driven models
benefit them. In a recent survey of the top ten highest selling for how different biomarkers, and possible even with different
drugs in the USA, it was reported that for each person benefitting quantitative levels, are associated with disease development
from any of these drugs, between 4 and 25 people are not being and thereby enable precision medicine, where the treatment is
helped (Schork, 2015). The healthcare sector is therefore in need tailored to deal with a specific molecular event underlying the
of transformation, both to reduce costs and to ensure better disease. There are already some ongoing N-of-1 studies where
treatment of patients. This requires that physicians consider healthy people are being monitored in detail over time, one being
the large variation between individuals to reach this objective. the so-called 100K project where the objective is to enroll
Most currently used pharmaceuticals have been developed 100,000 individuals and follow them longitudinally for 20 or
based on clinical trials involving large cohorts and are given to more years (Hood and Price, 2014). This project follows a
patients on the assumption that everyone will respond similarly. 9-month pilot study called the Hundred Person Wellness Project
This is neglecting the fact that there are large genetic and envi- (HPWP), where 100 individuals were intensively monitored and
ronmental differences between individuals, and recently it has offered regular feedback and counseling about lifestyle changes,
also been found that the gut microbiome has an influence on e.g., suggested dietary changes or altered sleeping habits
drug response, adding further complexity. In order to take these (Gibbs, 2014). In the HPWP, all individuals had their genome
variations into consideration, there is increasing interest in the sequenced at enrollment. Furthermore, insulin sensitivity, im-
concept of personalized medicine, which is based on stratifica- mune cell activity, 100 key proteins, and the gut microbiome
tion of patients into different molecularly defined groups and composition were monitored at enrollment and every 3 months.
then using different treatments and/or interventions for each Finally, pulse and sleeping patterns were monitored continu-
group (Figure 1A). With accumulating knowledge on the molec- ously using a wrist sensor. This resulted in the generation of a
ular mechanisms for many diseases and the development of very large amount of data for each individual, a virtual data cloud
more efficient diagnosis, there is increasing interest in moving consisting of billions of data points, and the challenge ahead is to
from disease treatment, the current practice, to disease preven- ensure efficient analysis of these data and extract information
tion, as this will significantly reduce costs in the healthcare that can be used for direct advice on lifestyle and/or treatment
sector. This, however, requires that identified biomarkers have strategies (Hood, 2013). Analysis of this kind of big data is chal-
truly predictive strength, something that can only be obtained lenging, as discussed later, but through the integration of the
through dedicated clinical studies, preferentially longitudinal data with mathematical models or reconstructed biological net-
over long time. Studies that focus on individuals, known as works, much new biological information can be derived. The use
N-of-1 trials (Figure 1B), are important for this, as these will pro- of computational and mathematical models for studying biolog-
vide data on variations within and between individuals and here- ical systems is referred to as systems biology, and when applied
by will enable the identification of which biomarkers that can be specifically for studying human diseases as systems medicine
used for solid stratification and for detection of disease onset. (Hood, 2013). Here I will discuss the challenges of systems
Often N-of-1 trials engage the patients actively, i.e., they become medicine and illustrate how one type of mathematical model,

572 Cell Metabolism 25, March 7, 2017 2017 Elsevier Inc.

Cell Metabolism

Perspective

Figure 1. Principles of Personalized Medicine

(A) Illustration of how analysis of big data obtained from detailed omics analysis of patient cohorts can result in detailed phenotyping and thereby lead to
stratification of patients into different groups. In connection with this, there can be identified a set of biomarkers that can be used for the stratification in the clinic.
These biomarkers are unique molecules (or combination of molecules), e.g., metabolites or proteins, that pass a certain level when specific cellular processes are
changed in connection with disease onset or progression.
(B) Illustration of the concept of N-of-1 trials. Each individual in the cohort is followed over time, during which samples are taken at different time points. This
detailed phenotyping over time enables identification of deviations from normal, which may point to disease development. Furthermore, N-of-1 trials will provide
information on variations of biomarkers both within and between individuals, and this will be important for identification of biomarkers that are truly predictive and
can therefore be used for stratification of cohorts not included in the N-of-1 study.

the so-called genome-scale metabolic model (GEM), can be cessed separately before they can be used for integrative analysis.
used as a scaffold for integrative analysis with the objective to Another challenge is that multi-omics data represent varying types
identify novel prognostic biomarkers that can assist in the of information with very different timescales and different dynamic
advancement toward personalized and precision medicine. ranges. Thus, metabolites change with completely different time
constants than mRNAs and proteins, and the level of metabolites
Challenges for Systems Medicine in a cell is determined not only by the enzyme levels, but also by the
Advancing systems medicine faces several challenges: (1) the kinetics of the individual enzymes; by post-translational modifica-
challenge of analyzing large datasets, (2) the difficulties in iden- tion of enzymes, e.g., protein phosphorylation and acetylation;
tifying mechanistic causes for many biomarkers and drug tar- and by metabolite regulation. Furthermore, metabolites circulating
gets, (3) problems with translation from model systems to the in the blood are determined not only by the metabolic activity of the
clinic, and (4) problems with sample heterogeneity. different tissues, but also by the food intake and by the metabolic
The detailed analysis underlying systems medicine results in activity of the human gut microbiota. Similarly, the plasma prote-
generation of very large datasets, generally referred to as big ome is a complex function of the physiological state of the different
data. Even though they are smaller in size than other types of big tissues and cell types in the body. It is therefore challenging to
data generated, e.g., in the financial sector, traffic control, and apply plasma metabolomics or proteomics for diagnosis and to
meteorology, it is challenging to analyze multiple types of omics integrate these data with other omics data, unless one uses a scaf-
data as there is a large variation in data structures and formats. fold that provides a priori information on how the different variables
Thus, a recent analysis demonstrated that with four different are connected.
data types, the resources required for data analysis are larger Metabolism is closely integrated with practically all cellular
than the resources for data generation for only four datasets, processes, and any kind of perturbation in cellular physiology
and the resource requirement for data analysis increases rapidly therefore typically results in an altered metabolic footprint, i.e.,
when more datasets are to be analyzed (Palsson and Zengler, altered uptake or secretion of metabolites from or to the blood.
2010). This is because different data types need to be pre-pro- Plasma metabolomics data therefore have a huge potential for

Cell Metabolism 25, March 7, 2017 573

 Cell Metabolism

Perspective

identification of altered health status. This concept is already metabolic changes associated with disease onset that is not
widely used for monitoring of triacylglycerides and cholesterols caused by genetic dispositions.
in the blood, but analysis of blood chemistry could probably be Another challenge for systems medicine is that even though
used much more widely for diagnosis. However, it is a major there are many good model systems available for studying
challenge to link an altered metabolite profile to onset or pro- different human diseases, the translation to the clinic often fails.
gression of a specific disease. To illustrate the complexity, the This is often ascribed to biological differences between, e.g.,
Human Metabolome Database (HMDB) includes about 42,000 mouse and human, but it may equally well be due to impacts
metabolites (Wishart et al., 2013; www.hmdb.ca). A large num- of diet and lifestyle, as well as the presence of much larger var-
ber of these are food metabolites (about 32,500) and drug me- iations in genetics and gut microbiome composition between hu-
tabolites (about 2,500), but still about 4,500 metabolites have man individuals in a clinical trial than in a controlled preclinical
been reported in serum (Psychogios et al., 2011; www.hmdb. study. In the field of cancer, complexity is further added by a
ca). With so many metabolites present in human serum, and large heterogeneity across and within tumors, which even ques-
the large sensitivity of the levels of many of these toward lifestyle tions the traditional histopathological classification of cancers.
differences, in particular diet, it is of course challenging to iden- N-of-1 trials on large cohorts will assist in overcoming some of
tify biomarkers associated with specific diseases solely from these challenges as it will allow the identification of the common-
human serum metabolome analysis. Only few biomarkers have alities across a population in connection with disease develop-
therefore been identified from this kind of analysis, exemplified ment, i.e., which are truly conserved biomarkers and associated
by the identification of elevated levels of branched-chain amino mechanisms, and which are associated with specific genetic dif-
acids as a marker for obesity and diabetes (Newgard et al., 2009; ferences and/or lifestyles. Such studies will therefore assist in the
Zhao et al., 2016). Even though transcriptome analysis of identification of prognostic biomarkers that can be used for strat-
abdominal human fat biopsies, enriched in adipocytes, showed ification and for prognosis of disease development.
that elevated levels of branched-chain amino acids in obese
subjects may be caused by reduced respiratory metabolism in The Central Role of Metabolism in Cellular Physiology
this tissue (Mardinoglu et al., 2014a), there is still lacking a mech- Metabolism plays a central role in living cells, for it provides the
anistic explanation for why these amino acids are such strong energy and building blocks for cellular growth as well as ensuring
biomarkers for obesity and diabetes. More detailed analysis of protection against external stress factors, e.g., xenobiotics and
metabolic alterations in different tissues is required for obtaining oxidative stress. Metabolism has evolved to support function
a mechanistic explanation for this finding, and this will require of the cell and can roughly be divided into three types: (1) central
both large datasets, e.g., transcriptome or proteome data, carbon metabolism, which ensures conversion of carbon and
from different tissues in large cohorts, and detailed models energy sources into free energy, redox power, and precursor
that can be used for integrative analysis of such data. The diffi- metabolites required for biosynthesis; (2) biosynthesis, where
culties with identification of prognostic biomarkers solely using precursors are converted into building blocks like amino acids,
plasma metabolomics are well illustrated by identification of sar- nucleotides, fatty acids, etc. required for cell growth; and (3) sec-
cosine for prostate cancer progression (Sreekumar et al., 2009). ondary and endogenous metabolism, which is typically highly
Later studies could not validate these findings (Jentzmik et al., diverse among cells. Enzymes of the central carbon metabolism
2011; Ankerst et al., 2015), and like many other biomarkers it are the most catalytically efficient but have evolved to generally
has therefore not been translated for clinical use. Through the be smaller than enzymes of other parts of metabolism (Bar-Even
combination of plasma metabolomics with other omics data, it et al., 2011). They are, however, still the most abundant in bacte-
is possible to get a mechanistic explanation for changes in ria, single-cell eukaryal cells like yeast, and human (Liebermeis-
metabolite levels. This has been demonstrated in several studies ter et al., 2014). In microbes, about half of the proteome is allo-
where plasma metabolomics was used for genome-wide associ- cated to metabolism, with about 25% being allocated alone to
ation studies (GWASs). Using GWASs of more than 200 metab- glycolysis, whereas in human this number is lower as a larger
olites in a large cohort of more than 2,000 subjects with a fraction of the proteome is allocated to cytoskeleton proteins,
detailed cardiometabolic phenotyping resulted in identification chaperones, and the spliceosome (Liebermeister et al., 2014).
of inborn mutations in AGXT2, a transaminase, being associated The high catalytic efficiency, small size, and high abundance of
with altered cholesterol and triacylglycerol levels (Rhee et al., enzymes in the central carbon metabolism are consistent with
2013). In a later study on the same cohort, but using more the central role this part of metabolism is playing in ensuring con-
detailed genetic profiling, mutations in several metabolic en- stant provision of energy, primarily in the form of ATP, in handling
zymes were identified to be associated with altered plasma electron flows by balancing the co-factors NADH and NADPH,
metabolite levels (Rhee et al., 2016). GWAS analysis has also and in providing precursors for cellular growth. Thus, the flux
been done with metabolome data from urine samples, and here- through the central carbon metabolism typically exceeds the
by several loci were identified that have also earlier been identi- flux through other metabolic pathways by a factor 10 or more.
fied to be important for clinical outcomes, and this led to the With these multiple roles, the central carbon metabolism has to
identification of several potential metabolite biomarkers that be highly connected with the other parts of metabolism, i.e., in
can be measured in urine (Suhre et al., 2011). These are exam- yeast ATP is used in more than 200 out of about 1,500 metabolic
ples of single genetic differences specifically causing an altered reactions, and metabolism therefore forms a highly connected
enzyme activity, and they are valuable for identification of pa- metabolic network (J.N., unpublished data). This means that a
tients with increased risks for disease development, but GWASs perturbation of almost any part of metabolism results in a global
of plasma metabolome data do not allow for gaining insight into response in which a large number of enzymes have to alter their

574 Cell Metabolism 25, March 7, 2017

Cell Metabolism

Perspective

B C

Figure 2. Illustration of the Concept of Integrative Data Analysis Using Metabolic Networks
(A) Illustration of how a metabolic map, represented by a genome-scale metabolic model (GEM), can be used for integrative analysis of omics data, e.g.,
transcriptome, proteome, or metabolome data. By overlaying these data on the metabolic map, it is possible to identify reporter metabolites and/or sub-networks
that represent parts of metabolism that have altered activity in response to change, e.g., disease development. A set of reporter metabolites may be connected in
the metabolic network and thereby point to altered activity of non-canonical pathways.
(B) Illustration of how tissue-specific models are a subset of a generic GEM for human metabolism, here illustrated by HMR2.
(C) Example of a reporter sub-network identified in ccRCC using a specific cancer GEM together with transcriptome data from both the cancer tissue and
corresponding healthy kidney tissue. The sub-network involves a large number of reactions in heparan and chondroitin sulfate biosynthesis pointing to altered
levels of metabolism in plasma and urine.

function in order to maintain homeostasis. This explains why Genome-scale Metabolic Models
almost any perturbation of cellular physiology will have a meta- Concept
bolic fingerprint, i.e., changes in a certain part of metabolism, GEMs are comprehensive compilations of all the metabolic reac-
and this may be quite specific. It further means that with the tions that take place in a particular cell, tissue, organ, or organism
high degree of connectivity in metabolism, it is difficult to analyze (OBrien et al., 2015). Each reaction is associated with one or more
changes in metabolism without the use of mathematical models. enzymes and encoded by specific genes; thus, a direct gene-pro-
I therefore hypothesize that any disease onset will result in a shift tein-reaction connection can be established. This is an important
in the metabolic homeostasis in the body, and such shifts can feature of GEMs as it allows for overlaying omics-type data, e.g.,
possibly be detected through metabolome analysis of plasma. transcriptome or proteome data, and thereby identifying co-regu-
These changes may be very small, in particular at the early stage lated sub-networks in metabolism (Figure 2A) (Patil and Nielsen,
of disease onset, and therefore difficult to detect unless a tar- 2005). These co-regulated sub-networks, or reporter metabolites,
geted approach is applied. This has to follow a hypothesis gener- point to parts of the metabolism that need to have altered expres-
ated from analysis of, e.g., transcriptome or proteome data from sion in order to maintain cellular homeostasis. Often these co-
tissues associated with the disease combined with integrative regulated sub-networks are not directly associated with the parts
analysis. As will be discussed below, GEMs represent an excel- of metabolism that are affected (Patil and Nielsen, 2005). Thus, if
lent scaffold for this kind of analysis. cells are exposed to oxidative stress there may be alterations

Cell Metabolism 25, March 7, 2017 575

 Cell Metabolism

Perspective

not only in glutathione metabolism that is directly engaged in basis for the elevated levels of these metabolites in plasma
coping with the oxidative stress, but also in more distant parts of in obese subjects or those with T2D. Other tissue-specific
metabolism, e.g., the pentose phosphate pathway, ensuring GEMs have also been reconstructed computationally using
regeneration of NADPH used in glutathione metabolism. data from tissue-specific gene expression values (Shlomi et al.,
Through specification of the stoichiometry of the different re- 2008) or from data from the Human Protein Atlas (HPA) (www.
actions in a metabolic network, GEMs can be used for simulation proteinatlas.org) (Agren et al., 2012, 2014). HPA data are partic-
of metabolic functions using the concept of flux balance analysis ularly well suited for the generation of cell-type-specific GEMs,
(OBrien et al., 2015). This concept assumes that all fluxes into for immunohistochemistry has been used for identifying the
a metabolite pool equal all fluxes out of the pool. Of course, presence of proteins in 80 different human cell types, and cell-
perturbations of metabolism will result in deviations from this type-specific models can therefore be generated. These models
steady-state condition, but the flux through most metabolite allow for direct analysis of the metabolism of different cell types
pools is so high that the pool turnover is on the order of seconds present in tissues, and thereby enable better understanding of
or minutes (depending on the part of metabolism), meaning the mechanisms underlying changes in overall tissue meta-
that a deviation from flux balancing will be resolved in just a bolism. RNA sequencing (RNA-seq) has recently been shown
few seconds/minutes by the resulting rapid change in metabolite to provide much new insight into biological differences between
levels. Flux balance analysis imposes a large number of different human tissues, and using this kind of data 32 tissue-
constraints on the fluxes and thereby allows for calculation of specific GEMs were generated (Uhlen et al., 2015). Human
fluxes through different parts of the metabolism based on mea- GEMs have also been used for the identification of novel
surements of a few exchange fluxes, e.g., fluxes of nutrient up- drug targets for cancer treatment (Folger et al., 2011), as thor-
take, but as the degrees of freedom in these models is quite oughly reviewed elsewhere (Yizhak et al., 2015), and recently
large, all fluxes cannot be uniquely determined (Mardinoglu illustrated for argininosuccinate synthase (ASS1)-deficient tu-
and Nielsen, 2015). Recently it has, however, been shown that mors (Rabinovich et al., 2015). These tumors have elevated
by incorporating kinetic information into GEMs, together with a levels of aspartate, which is beneficial for de novo pyrimidine
constraint on proteome usage for metabolic enzymes, it is biosynthesis, and it is therefore important to block this part of
possible to improve the predictive strength of GEMs significantly metabolism in ASS1-deficient tumors. As mentioned above,
(Thiele et al., 2012; Nilsson and Nielsen, 2016) and thereby cancer cells are extremely heterogeneous, and using proteomics
describe overflow metabolism to lactate in cancer cells (Shlomi data from hepatocellular carcinoma (HCC) tumors, personalized
et al., 2011). GEMs were generated for six individuals with HCC (Agren et al.,
Human GEMs 2014). HCC metabolism was indeed found to be quite different in
In 2007, the two first GEMs for human metabolism were recon- the six individuals, but by using the GEMs it was possible to iden-
structed (Ma et al., 2007; Duarte et al., 2007), and these models tify anti-metabolites that block cell growth in all six tumors. One
formed the basis for Recon2, a much expanded model with of these targets was the carnitine carrier system, which is
broader coverage of metabolism (Thiele et al., 2013). In connec- responsible for the transport of fatty acids into the mitochondria
tion with building tissue-specific GEMs, more details in lipid for b-oxidation and thereby ensures sufficient energy generation
metabolism had to be incorporated and this resulted in Human for the cancer cells. Using HepG2 cells, a cell line derived from
Metabolic Reaction (HMR2) (Agren et al., 2014), which is HCC tumors, this target was validated and shown to prevent
currently the most comprehensive GEM for human cells, cell proliferation (Agren et al., 2014). Considering the large het-
covering 3,765 genes, 8,181 reactions, and 6,007 metabolites. erogeneity in the six tumors, it is, however, very likely that this
HMR2 has been used as a basis for reconstruction of detailed identified drug target may not constitute an effective treatment
models for different human cell types, which become sub-sets across larger cohorts, clearly pointing to the need for a more
of HMR2 (Figure 2B). Cell-type-specific GEMs have been recon- personalized approach to cancer treatment. GEMs were also
structed for adipocytes (Mardinoglu et al., 2013), hepatocytes used to contextualize gene expression changes independently
(Mardinoglu et al., 2014b), and myocytes (Va remo et al., 2015). associated with distinct cancer mutations and revealed a trans-
The adipocyte model was used for integrative analysis with the versal metabolic signature revolving around arachidonic acid
objective of gaining insight into metabolic reprogramming in and xenobiotic metabolism (Gatto et al., 2016a). This finding
abdominal fat tissues in response to obesity, and it was found may be important as it could lead to the identification of a treat-
that respiratory metabolism was significantly reduced in obese ment strategy that can be used for several cancer types.
subjects. At the same time, catabolism of branched-chain amino Identification of Metabolite Biomarkers
acids (valine, leucine, and isoleucine) was found to be attenuated GEMs have in several cases demonstrated their power for iden-
(Mardinoglu et al., 2014a), which can explain the elevated levels tification of biomarkers that have subsequently been validated
of these metabolites in plasma (Newgard et al., 2009). The adipo- from plasma metabolomics. Using a hepatocyte GEM, it was
cyte model was also used to illustrate that attenuated respiration possible to study metabolic reprogramming in response to
caused problems with oxidation of accumulated triacylglycerols development of non-alcoholic fatty liver disease (NAFLD) (Mardi-
and therefore resulted in reduced dynamics of lipid bodies in noglu et al., 2014b). From this analysis, it was found that patients
obese subjects (Mardinoglu et al., 2013). The myocyte model developing non-alcoholic steatohepatitis (NASH) had a signifi-
was similarly used to identify co-regulated networks in meta- cant decreased expression of genes encoding for enzymes in
bolism in response to type 2 diabetes (T2D), and for muscle tis- serine and glycine biosynthesis, which can explain observation
sue attenuated catabolism of branched-chain amino acids was of elevated levels of plasma homocysteine (Gulsen et al., 2005)
remo et al., 2015), further pointing to a mechanistic
identified (Va and decreased levels of phosphatidylserine in the liver of

576 Cell Metabolism 25, March 7, 2017

Cell Metabolism

Perspective

NASH patients (Gorden et al., 2011). This finding was validated in ongoing and planned N-of-1 clinical trials, it is important to
a follow-up study in which it was shown that NASH patients have further expand and include more subjects and also expand the
reduced levels of serine and glycine in the plasma, pointing to scope of some of these studies to ensure that very detailed
serine deficiency in these patients (Mardinoglu et al., 2016). phenotypic characterization of the individuals is performed. As
Moreover, serine supplementation could improve the health sta- discussed, GEMs offer much in terms of integrative analysis,
tus of such patients. This study gives a very strong indication that and through further expansion of the models with description
serine and glycine levels in plasma can be used as a non-invasive of protein synthesis and other cellular processes, the scope of
biomarker for NASH development in patients with a fatty liver. these models will expand and allow for simulating the impact
HMR2 has also been used to find a very strong prognostic of many key cellular processes underlying human diseases,
biomarker for clear cell renal cell carcinoma (ccRCC). This was e.g., oxidative stress and protein mis-folding stress. Other
identified from a study that initially evaluated metabolic reprog- computational approaches should, however, also be consid-
ramming in eight different cancers using RNA-seq data from ered. Recent development in machine learning, with emphasis
the Cancer Genome Atlas (TGCA) database (Gatto et al., on deep learning (Angermueller et al., 2016), has shown to be
2014). From this analysis, ccRCC was found to have a unique powerful for analyzing large datasets and holds promise to adapt
metabolic reprograming, distinctive from the other epithelial can- to problems in computational biology that may in the future
cers. This was, in turn, associated with repression of metabolic assist with diagnostics in the clinic. This was excellently illus-
functions in several different parts of metabolism, e.g., nucleo- trated in a large dietary N-of-1 clinical study that was carried
tide metabolism, which makes the tumor more vulnerable out with the objective of enabling personalized dietary advice
against inhibition of specific enzymatic functions according to (Zeevi et al., 2015). Using a very large dataset, involving an
experimentally validated GEM-based simulations (Gatto et al., 800-person cohort with measured responses to more than
2015). More importantly, the integrative data analysis also iden- 45,000 meals, a machine-learning algorithm was generated by
tified a strong de-regulation of heparan and chondroitin sulfate integrating blood chemistry, dietary habits, and gut microbiota
biosynthesis, and subsequent quantification of these metabo- composition. Using the algorithm, it was possible to successfully
lites in the plasma and urine of patients with metastatic ccRCC predict glycemic responses in a 100-person follow-up cohort,
resulted in identification of a systems biomarker that is deter- demonstrating that this algorithm can be used for personalized
mined by altered levels of several of these metabolites (Gatto nutritional advice. Even though machine-learning algorithms
et al., 2016b). This systems biomarker was further found to cannot directly provide mechanistic insight, these algorithms still
have prognostic value; it can predict the aggressiveness of the allow for providing clear connectivity between a very large num-
tumor and thereby survival rate of ccRCC patients (Gatto et al., ber of variables, and these can then be used for follow-up studies
2016c), and it is now being brought to the clinic for evaluation with the objective of identifying the underlying mechanisms.
of its diagnostic and predictive capabilities for the treatment The above-mentioned study, like many other N-of-1 clinical tri-
of ccRCC. als, included analysis of the gut microbiota, as this has been
Finally, a recent study used HMR2 in combination with a bio- shown to have a large impact on overall human metabolism
logical network derived from protein-protein interactions for (Karlsson et al., 2013a; Arora and Ba ckhed, 2016; Wu et al.,
analysis of transcriptome and proteome data for insulin-resistant 2015). However, even though clear correlations have been iden-
patients and matched controls (Lee et al., 2016). This resulted in tified between the gut microbiota and many different human dis-
the identification of mannose metabolism to be significantly eases, e.g., T2D (Karlsson et al., 2013b), most of these studies
altered in insulin-resistant patients, and subsequent analysis of are only correlative and no causal effects have been identified.
metabolomics from more than 1,000 subjects could validate Here mathematical modeling can assist in gaining insight into
mannose as a novel biomarker for insulin resistance (Lee the interaction between the many different species and their
et al., 2016). host (Heinken and Thiele, 2015). The gut microbiota represents
The above-mentioned studies are all examples of how sys- a very complex ecosystem with a large number of species that
tems biology analysis of specific human tissues resulted in the express different metabolic phenotypes. GEMs are well suited
identification of changes in specific parts of the metabolic for modeling of this kind of ecosystem: models for individual spe-
network, and these changes resulted in altered plasma metabo- cies can capture the overall metabolism of each species, and
lite levels. It would have been difficult to identify these bio- various algorithms can then be used for simulation of their inter-
markers without a directed search, but based on identified and actions (Shoaie et al., 2013). Hereby it has been demonstrated
statistically significant alterations in the metabolic networks, a that it is possible to simulate how the human gut microbiota is
hypothesis could be generated about certain metabolites being impacted by diet and how it impacts plasma chemistry, including
likely biomarkers, and from targeted metabolomics these could the level of many amino acids (Shoaie et al., 2015). Even though
thereafter be validated. The strength of this approach is that this last study only considered the five most dominant species in
not only are novel biomarkers identified, but a mechanistic the gut microbiota, it clearly demonstrates that it is becoming
explanation for their function is directly provided. possible to simulate how this complex ecosystem is impacted
by diet and how it interacts with host metabolism. By adding
Perspectives more models, it will become possible to simulate not only the
There are some challenges for advancing systems medicine, impact of diet on the gut microbiome development but also
but these basically condense into developing better methods how the gut microbiome should be modulated, e.g., through
for integrative analysis of data and the establishment of N-of-1 addition of new probiotics, in order to attain properties associ-
clinical trials with large cohorts. Even though there are several ated with healthy subjects. Here a recent study describing 773

Cell Metabolism 25, March 7, 2017 577

 Cell Metabolism

Perspective

GEMs for gut symbionts provides a valuable resource for ex- Gatto, F., Schulze, A., and Nielsen, J. (2016a). Systematic analysis reveals that
cancer mutations converge on deregulated metabolism of arachidonate and
panding our description of the gut microbiota metabolism xenobiotics. Cell Rep. 16, 878895.
(Magnusdottir et al., 2017). Hereby it may also become possible
to use probiotics as combination treatment with drugs that are Gatto, F., Volpi, N., Nilsson, H., Nookaew, I., Maruzzo, M., Roma, A., Johans-
son, M.E., Stierner, U., Lundstam, S., Basso, U., and Nielsen, J. (2016b).
impacted by the gut microbiota composition, as identified for Glycosaminoglycan profiling in patients plasma and urine predicts the occur-
some anti-cancer drugs (Vetizou et al., 2015; Sivan et al., 2015). rence of metastatic clear cell renal cell carcinoma. Cell Rep. 15, 18221836.
From the above, it is clear that systems biology can lead to Gatto, F., Maruzzo, M., Magro, C., Basso, U., and Nielsen, J. (2016c). Prog-
identification of novel biomarkers and drug targets, and at the nostic value of plasma and urine glycosaminoglycan scores in clear cell renal
same time provide a mechanistic explanation for why they can cell carcinoma. Front. Oncol. 6, 253.
be used for diagnosis and in development of effective treatment Gibbs, W.W. (2014). Medicine gets up close and personal. Nature 506,
strategies. However, much more data are needed in order to 144145.
develop strong biomarkers that are personalized and allow for Gorden, D.L., Ivanova, P.T., Myers, D.S., McIntyre, J.O., VanSaun, M.N.,
precise detection of disease onset. GEMs represent an excellent Wright, J.K., Matrisian, L.M., and Brown, H.A. (2011). Increased diacylglycer-
scaffold for analysis of this kind of data, and a particular strength ols characterize hepatic lipid changes in progression of human nonalcoholic
fatty liver disease; comparison to a murine model. PLoS ONE 6, e22775.
of these models is that they are open ended in the sense that they
can be expanded with more biological knowledge and thereby Gulsen, M., Yesilova, Z., Bagci, S., Uygun, A., Ozcan, A., Ercin, C.N., Erdil, A.,
acquire increasing predictive strength. I am therefore confident Sanisoglu, S.Y., Cakir, E., Ates, Y., et al. (2005). Elevated plasma homocyste-
ine concentrations as a predictor of steatohepatitis in patients with non-alco-
that together with big data obtained from large N-of-1 clinical holic fatty liver disease. J. Gastroenterol. Hepatol. 20, 14481455.
studies, GEMs will contribute significantly to the advancement
Heinken, A., and Thiele, I. (2015). Systematic prediction of health-relevant
of personalized and precision medicine in the next 510 years. human-microbial co-metabolism through a computational framework. Gut
Microbes 6, 120130.
ACKNOWLEDGMENTS Hood, L. (2013). Systems biology and p4 medicine: past, present, and future.
Rambam Maimonides Med. J. 4, e0012.
I would like to acknowledge valuable discussions with Adil Mardinoglu, Fran-
cesco Gatto, and Jon Robinson. I also thank Francesco Gatto with assistance Hood, L., and Friend, S.H. (2011). Predictive, personalized, preventive, partic-
in drafting the figures. I acknowledge funding to my lab from Knut and Alice ipatory (P4) cancer medicine. Nat. Rev. Clin. Oncol. 8, 184187.
Wallenberg Foundation, the Novo Nordisk Foundation, Vetenskapsradet, Bill
Hood, L., and Price, N.D. (2014). Demystifying disease, democratizing health
& Melinda Gates Foundation, FORMAS, and the Swedish Foundation for Stra-
care. Sci. Transl. Med. 6, 225ed5.
tegic Research.
Jentzmik, F., Stephan, C., Lein, M., Miller, K., Kamlage, B., Bethan, B., Kris-
REFERENCES tiansen, G., and Jung, K. (2011). Sarcosine in prostate cancer tissue is not a
differential metabolite for prostate cancer aggressiveness and biochemical
progression. J. Urol. 185, 706711.
Agren, R., Bordel, S., Mardinoglu, A., Pornputtapong, N., Nookaew, I., and
Nielsen, J. (2012). Reconstruction of genome-scale active metabolic networks ckhed, F. (2013a). Assessing the
Karlsson, F., Tremaroli, V., Nielsen, J., and Ba
for 69 human cell types and 16 cancer types using INIT. PLoS Comput. Biol. 8, human gut microbiota in metabolic diseases. Diabetes 62, 33413349.
e1002518.
Karlsson, F.H., Tremaroli, V., Nookaew, I., Bergstrom, G., Behre, C.J., Fager-
Agren, R., Mardinoglu, A., Asplund, A., Kampf, C., Uhlen, M., and Nielsen, J. ckhed, F. (2013b). Gut metagenome in European
berg, B., Nielsen, J., and Ba
(2014). Identification of anticancer drugs for hepatocellular carcinoma through women with normal, impaired and diabetic glucose control. Nature
personalized genome-scale metabolic modeling. Mol. Syst. Biol. 10, 721. 498, 99103.

rnamaa, T., Parts, L., and Stegle, O. (2016). Deep learning

Angermueller, C., Pa Lee, S., Zhang, C., Kilicarslan, M., Piening, B.D., Bjornson, E., Hallstrom, B.M.,
for computational biology. Mol. Syst. Biol. 12, 878. Groen, A.K., Ferrannini, E., Laakso, M., Snyder, M., et al. (2016). Integrated
network analysis reveals an association between plasma mannose levels
Ankerst, D.P., Liss, M., Zapata, D., Hoefler, J., Thompson, I.M., and Leach, and insulin resistance. Cell Metab. 24, 172184.
R.J. (2015). A case control study of sarcosine as an early prostate cancer
detection biomarker. BMC Urol. 15, 99. Liebermeister, W., Noor, E., Flamholz, A., Davidi, D., Bernhardt, J., and Milo, R.
(2014). Visual account of protein investment in cellular functions. Proc. Natl.
ckhed, F. (2016). The gut microbiota and metabolic disease:
Arora, T., and Ba Acad. Sci. USA 111, 84888493.
current understanding and future perspectives. J. Intern. Med. 280, 339349.
Ma, H., Sorokin, A., Mazein, A., Selkov, A., Selkov, E., Demin, O., and Gorya-
Bar-Even, A., Noor, E., Savir, Y., Liebermeister, W., Davidi, D., Tawfik, D.S., nin, I. (2007). The Edinburgh human metabolic network reconstruction and its
and Milo, R. (2011). The moderately efficient enzyme: evolutionary and phys- functional analysis. Mol. Syst. Biol. 3, 135.
icochemical trends shaping enzyme parameters. Biochemistry 50, 44024410.
Magnusdottir, S., Heinken, A., Kutt, L., Ravcheev, D.A., Bauer, E., Noronha, A.,
Duarte, N.C., Becker, S.A., Jamshidi, N., Thiele, I., Mo, M.L., Vo, T.D., Srivas, Greenhalgh, K., Ja ger, C., Baginska, J., Wilmes, P., et al. (2017). Generation of
R., and Palsson, B.. (2007). Global reconstruction of the human metabolic genome-scale metabolic reconstructions for 773 members of the human gut
network based on genomic and bibliomic data. Proc. Natl. Acad. Sci. USA microbiota. Nat. Biotechnol. 35, 8189.
104, 17771782.
Mardinoglu, A., and Nielsen, J. (2015). New paradigms for metabolic modeling
Folger, O., Jerby, L., Frezza, C., Gottlieb, E., Ruppin, E., and Shlomi, T. (2011). of human cells. Curr. Opin. Biotechnol. 34, 9197.
Predicting selective drug targets in cancer through metabolic networks. Mol.
Syst. Biol. 7, 501. Mardinoglu, A., Agren, R., Kampf, C., Asplund, A., Nookaew, I., Jacobson, P.,
Walley, A.J., Froguel, P., Carlsson, L.M., Uhlen, M., and Nielsen, J. (2013). Inte-
Gatto, F., Nookaew, I., and Nielsen, J. (2014). Chromosome 3p loss of hetero- gration of clinical data with a genome-scale metabolic model of the human
zygosity is associated with a unique metabolic network in clear cell renal car- adipocyte. Mol. Syst. Biol. 9, 649.
cinoma. Proc. Natl. Acad. Sci. USA 111, E866E875.
Mardinoglu, A., Kampf, C., Asplund, A., Fagerberg, L., Hallstrom, B.M., Ed-
Gatto, F., Miess, H., Schulze, A., and Nielsen, J. (2015). Flux balance analysis her, M., Ponten, F., Uhlen, M., and Nielsen, J. (2014a). Defining
lund, K., Blu
predicts essential genes in clear cell renal cell carcinoma metabolism. Sci. the human adipose tissue proteome to reveal metabolic alterations in obesity.
Rep. 5, 10738. J. Proteome Res. 13, 51065119.

578 Cell Metabolism 25, March 7, 2017

Cell Metabolism

Perspective
Mardinoglu, A., Agren, R., Kampf, C., Asplund, A., Uhlen, M., and Nielsen, J. Shoaie, S., Ghaffari, P., Kovatcheva-Datchary, P., Mardinoglu, A., Sen, P.,
(2014b). Genome-scale metabolic modelling of hepatocytes reveals serine Pujos-Guillot, E., de Wouters, T., Juste, C., Rizkalla, S., Chilloux, J., et al.;
deficiency in patients with non-alcoholic fatty liver disease. Nat. Commun. MICRO-Obes Consortium (2015). Quantifying diet-induced metabolic changes
5, 3083. of the human gut microbiome. Cell Metab. 22, 320331.

Mardinoglu, A., Bjornson, E., Zhang, C., Adiels, M., Barrett, P.H.R., et al. Sivan, A., Corrales, L., Hubert, N., Williams, J.B., Aquino-Michaels, K., Earley,
(2016). Personal model-assisted identification of NAD+ and glutathione meta- Z.M., Benyamin, F.W., Lei, Y.M., Jabri, B., Alegre, M.L., et al. (2015).
bolism as intervention target in NAFLD. Mol. Syst. Biol. Published online March Commensal Bifidobacterium promotes antitumor immunity and facilitates
2, 2017. http://dx.doi.org/10.15252/msb.20167422. anti-PD-L1 efficacy. Science 350, 10841089.

Newgard, C.B., An, J., Bain, J.R., Muehlbauer, M.J., Stevens, R.D., Lien, L.F., Sreekumar, A., Poisson, L.M., Rajendiran, T.M., Khan, A.P., Cao, Q., Yu, J.,
Haqq, A.M., Shah, S.H., Arlotto, M., Slentz, C.A., et al. (2009). A branched- Laxman, B., Mehra, R., Lonigro, R.J., Li, Y., et al. (2009). Metabolomic profiles
chain amino acid-related metabolic signature that differentiates obese and delineate potential role for sarcosine in prostate cancer progression. Nature
lean humans and contributes to insulin resistance. Cell Metab. 9, 311326. 457, 910914.

Nilsson, A., and Nielsen, J. (2016). Metabolic trade-offs in yeast are caused by Suhre, K., Wallaschofski, H., Raffler, J., Friedrich, N., Haring, R., Michael, K.,
F1F0-ATP synthase. Sci. Rep. 6, 22264. Wasner, C., Krebs, A., Kronenberg, F., Chang, D., et al. (2011). A genome-
wide association study of metabolic traits in human urine. Nat. Genet. 43,
OBrien, E.J., Monk, J.M., and Palsson, B.O. (2015). Using genome-scale 565569.
models to predict biological capabilities. Cell 161, 971987.
Thiele, I., Fleming, R.M., Que, R., Bordbar, A., Diep, D., and Palsson, B.O.
Palsson, B., and Zengler, K. (2010). The challenges of integrating multi-omic (2012). Multiscale modeling of metabolism and macromolecular synthesis in
data sets. Nat. Chem. Biol. 6, 787789. E. coli and its application to the evolution of codon usage. PLoS ONE 7,
e45635.
Patil, K.R., and Nielsen, J. (2005). Uncovering transcriptional regulation of
metabolism by using metabolic network topology. Proc. Natl. Acad. Sci. Thiele, I., Swainston, N., Fleming, R.M.T., Hoppe, A., Sahoo, S., Aurich, M.K.,
USA 102, 26852689. Haraldsdottir, H., Mo, M.L., Rolfsson, O., Stobbe, M.D., et al. (2013). A com-
munity-driven global reconstruction of human metabolism. Nat. Biotechnol.
Pfuntner, A., Wier, L.M., and Elixhauser, A. (2011). Overview of Hospital Stays 31, 419425.
in the United States. HCUP Statistical Brief #166 (Rockville, USA: Agency for
Healthcare Research and Quality). Uhlen, M., Fagerberg, L., Hallstrom, B.M., Lindskog, C., Oksvold, P., Mardino-
glu, A., Sivertsson, A., Kampf, C., Sjostedt, E., Asplund, A., et al. (2015). Pro-
Psychogios, N., Hau, D.D., Peng, J., Guo, A.C., Mandal, R., Bouatra, S., Sinel- teomics. Tissue-based map of the human proteome. Science 347, 1260419.
nikov, I., Krishnamurthy, R., Eisner, R., Gautam, B., et al. (2011). The human remo, L., Scheele, C., Broholm, C., Mardinoglu, A., Kampf, C., Asplund, A.,
Va
serum metabolome. PLoS ONE 6, e16957.
Nookaew, I., Uhlen, M., Pedersen, B.K., and Nielsen, J. (2015). Proteome- and
transcriptome-driven reconstruction of the human myocyte metabolic network
Rabinovich, S., Adler, L., Yizhak, K., Sarver, A., Silberman, A., Agron, S., Stett-
and its use for identification of markers for diabetes. Cell Rep. 11, 921933.
ner, N., Sun, Q., Brandis, A., Helbling, D., et al. (2015). Diversion of aspartate in
ASS1-deficient tumours fosters de novo pyrimidine synthesis. Nature 527, Vetizou, M., Pitt, J.M., Daillere, R., Lepage, P., Waldschmitt, N., Flament, C.,
379383. Rusakiewicz, S., Routy, B., Roberti, M.P., Duong, C.P., et al. (2015). Anticancer
immunotherapy by CTLA-4 blockade relies on the gut microbiota. Science
Rhee, E.P., Ho, J.E., Chen, M.H., Shen, D., Cheng, S., Larson, M.G., Ghorbani, 350, 10791084.
A., Shi, X., Helenius, I.T., ODonnell, C.J., et al. (2013). A genome-wide asso-
ciation study of the human metabolome in a community-based cohort. Cell WHO (2011). World Health Statistics 2011 (Geneva, Switzerland: World Health
Metab. 18, 130143. Organization).
Rhee, E.P., Yang, Q., Yu, B., Liu, X., Cheng, S., Deik, A., Pierce, K.A., Bullock, Wishart, D.S., Jewison, T., Guo, A.C., Wilson, M., Knox, C., Liu, Y., Djoumbou,
K., Ho, J.E., Levy, D., et al. (2016). An exome array study of the plasma metab- Y., Mandal, R., Aziat, F., Dong, E., et al. (2013). HMDB 3.0The Human
olome. Nat. Commun. 7, 12360. Metabolome Database in 2013. Nucleic Acids Res. 41, D801D807.
Schork, N.J. (2015). Personalized medicine: time for one-person trials. Nature ckhed, F. (2015). Linking microbiota to human
Wu, H., Tremaroli, V., and Ba
520, 609611. diseases: a systems biology perspective. Trends Endocrinol. Metab. 26,
758770.
Shlomi, T., Cabili, M.N., Herrgard, M.J., Palsson, B.O., and Ruppin, E. (2008).
Network-based prediction of human tissue-specific metabolism. Nat. Bio- Yizhak, K., Chaneton, B., Gottlieb, E., and Ruppin, E. (2015). Modeling cancer
technol. 26, 10031010. metabolism on a genome scale. Mol. Syst. Biol. 11, 817.

Shlomi, T., Benyamini, T., Gottlieb, E., Sharan, R., and Ruppin, E. (2011). Zeevi, D., Korem, T., Zmora, N., Israeli, D., Rothschild, D., Weinberger, A., Ben-
Genome-scale metabolic modeling elucidates the role of proliferative adapta- Yacov, O., Lador, D., Avnit-Sagi, T., Lotan-Pompan, M., et al. (2015). Personal-
tion in causing the Warburg effect. PLoS Comput. Biol. 7, e1002018. ized nutrition by prediction of glycemic responses. Cell 163, 10791094.

Shoaie, S., Karlsson, F., Mardinoglu, A., Nookaew, I., Bordel, S., and Nielsen, Zhao, X., Han, Q., Liu, Y., Sun, C., Gang, X., and Wang, G. (2016). The relation-
J. (2013). Understanding the interactions between bacteria in the human gut ship between branched-chain amino acid related metabolomics signature and
through metabolic modeling. Sci. Rep. 3, 2532. insulin resistance: A systematic review. J. Diabetes Res. 2016, 112.

Cell Metabolism 25, March 7, 2017 579