Leaders in Pharmaceutical Business Intelligence (LPBI) Group

Prefacing the e-Book Epilogue: Metabolic Genomics and Pharmaceutics

Author and Curator: Larry H. Bernstein, MD, FCAP

 

Adieu, adieu, adieu …

Sound of Music

This work has been a coming to terms with my scientific and medical end of career balancing in a difficult time after retiring, but it has been rewarding.  In the clinical laboratories, radiology, anesthesiology, and in pharmacy, there has been some significant progress in support of surgical, gynecological, developmental, medical practices, and even neuroscience directed disciplines, as well as epidemiology over a period of half a century.  Even then, cancer and neurological diseases have been most difficult because the scientific basic research has either not yet uncovered a framework, or because that framework has proved to be multidimensional.  In the clinical laboratory sciences, there has been enormous progress in instrumental analysis, with the recent opening of molecular methods not yet prepared for routine clinical use, which will be a very great challenge to the profession, which has seen the development of large sample volume, multianalite, high-throughput, low-cost support emerging for decades.  The capabilities now underway will also enrrich the the capabilities of the anatomic pathology suite and the capabilities of 3-dimensional radiological examination.  In both pathology and radiology, we have seen the division of the fields into major subspecialties.  The development of the electronic health record had to take lessons from the first developments in the separate developments of laboratory, radiology, and pharmacy health record systems, to which were added, full cardiology monitoring systems.  These have been unintegrated.  This made it difficult to bring forth a suitable patient health record because the information needed to support decision-making by practitioners was in separate “silos”.  The mathematical methods that are being applied to the -OMICS sciences, can be brought to bear on the simplification and amplification of the clinicians’ ability to make decisions with near “errorless” discrimination, still allowing for an element of “art” in carrying out the history, physical examination, and knowledge unique to every patient.

We are at this time opening a very large, complex, study of biology in relationship to the human condition.  This will require sufficient resources to be invested in the development of these for a better society, which I suspect, will go on beyond the life of my grandchildren.  Hopefully, the long-term dangers of climate change will be controlled in that time.  As a society, or as a group of interdependent societies, we have no long term interest in continuing self-destructive behaviors that have predominated in the history of mankind.  I now top off these discussions with some further elucidation of what lies before us.

Metabolomics and systems pharmacology: why and how to model the human metabolic network for drug discovery

Douglas B. Kell and Royston Goodacre
School of Chemistry and Manchester Institute of Biotechnology, The University of Manchester, Manchester, UK
Drug Discovery Today Feb 2014;19(2)  http://dx.doi.org/10.1016/j.drudis.2013.07.014

Metabolism represents the ‘sharp end’ of systems biology,

• because changes in metabolite concentrations
• are necessarily amplified relative to
• changes in the transcriptome, proteome and enzyme activities,
• which can be modulated by drugs.

To understand such behaviour, we therefore need
(and increasingly have)

• reliable consensus (community) models of the human metabolic network
• that include the important transporters.

Small molecule ‘drug’ transporters are in fact metabolite transporters,

• because drugs bear structural similarities to metabolites known
• from the network reconstructions and from measurements of the metabolome.

Recon2 represents the present state-of-the-art human metabolic
network reconstruction; it can predict inter alia:

1. the effects of inborn errors of metabolism;
2. which metabolites are exometabolites, and
3. how metabolism varies between tissues and cellular compartments.

Even these qualitative network models are not yet complete. As our
understanding improves so do we recognize more clearly the need for a systems (poly)pharmacology.

Modelling biochemical networks – why we do so
There are at least four types of reasons as to why one would wish to model a biochemical network:

1. Assessing whether the model is accurate, in the sense that it
   reflects – or can be made to reflect – known experimental facts.
2. Establishing what changes in the model would improve the
   consistency of its behaviour with experimental observations
   and improved predictability, such as with respect to metabolite
   concentrations or fluxes.
3. Analyzing the model, typically by some form of sensitivity
   analysis, to understand which parts of the system contribute
   most to some desired functional properties of interest.
4. Hypothesis generation and testing, enabling one to analyse
   rapidly the effects of manipulating experimental conditions in
   the model without having to perform complex and costly
   experiments (or to restrict the number that are performed).

In particular, it is normally considerably cheaper to perform
studies of metabolic networks in silico before trying a smaller
number of possibilities experimentally; indeed for combinatorial
reasons it is often the only approach possible. Although
our focus here is on drug discovery, similar principles apply to the
modification of biochemical networks for purposes of ‘industrial’
or ‘white’ biotechnology.
Why we choose to model metabolic networks more than

• transcriptomic or proteomic networks

comes from the recognition – made particularly clear

• by workers in the field of metabolic control analysis

– that, although changes in the activities of individual enzymes tend to have

• rather small effects on metabolic fluxes,
• they can and do have very large effects on metabolite concentrations (i.e. the metabolome).

Modelling biochemical networks – how we do so

Although one could seek to understand the

1. time-dependent spatial distribution of signalling and metabolic substances within indivi
   dual cellular compartments and
2. while spatially discriminating analytical methods such as Raman spectroscopy and
   mass spectrometry do exist for the analysis of drugs in situ,

• the commonest type of modelling, as in the spread of substances in
  ecosystems,
• assumes ‘fully mixed’ compartments and thus ‘pools’ of metabolites.

Although an approximation, this ‘bulk’ modelling will be necessary for complex ecosystems such as humans where, in addition to the need for tissue- and cell-specific models, microbial communities inhabit this superorganism and the
gut serves as a source for nutrients courtesy of these symbionts.

Topology and stoichiometry of metabolic networks as major constraints on fluxes
Given their topology, which admits a wide range of parameters for
delivering the same output effects and thereby reflects biological
robustness,

• metabolic networks have two especially important constraints that assist their accurate modelling:

(i) the conservation of mass and charge, and
(ii) stoichiometric and thermodynamic constraints.

These are tighter constraints than apply to signalling networks.

New developments in modelling the human metabolic network
Since 2007, several groups have been developing improved but nonidentical models of the human metabolic network at a generalised level and in tissue-specific forms. Following a similar community-driven strategy in Saccharomyces cerevisiae, surprisingly similar to humans, and in Salmonella typhimurium,

we focus in particular on a recent consensus paper that provides a highly curated and semantically annotated model of the human metabolic network, termed

• Recon2 (http://humanmetabolism.org/).

In this work, a substantial number of the major groups active in this area came together to provide a carefully and manually constructed/curated network, consisting of some 1789 enzyme-encoding genes, 7440 reactions and 2626 unique metabolites distributed over eight cellular compartments.  A variety of dead-end metabolites and blocked reactions remain (essentially orphans and widows). But Recon2 was able to

• account for some 235 inborn errors of metabolism,
• a variety of metabolic ‘tasks’ (defined as a non-zero flux through a reaction or through a pathway leading to the production of a metabolite Q from a metabolite P).
• filtering based on expression profiling allowed the construction of 65 cell-type-specific models.
• Excreted or exometabolites are an interesting set of metabolites,
• and Recon2 could predict successfully a substantial fraction of those

Role of transporters in metabolic fluxes

The uptake and excretion of metabolites between cells and their macrocompartments

• requires specific transporters and in the order of one third of ‘metabolic’ enzymes,
• and indeed of membrane proteins, are in fact transporters or equivalent.

What is of particular interest (to drug discovery), based on their structural similarities, is the increasing recognition (Fig. 3) that pharmaceutical drugs also

• get into and out of cells by ‘hitchhiking’ on such transporters, and not –

to any significant extent –

• by passing through phospholipid bilayer portions
  of cellular membranes.

This makes drug discovery even more a problem of systems biology than of biophysics.

Two views of the role of solute carriers and other transporters in cellular drug uptake. (a) A more traditional view in which all so-called ‘passive’drug uptake occurs through any unperturbed bilayer portion of membrane that might be present.
(b) A view in which the overwhelming fraction of drug is taken up via solute transporters or other carriers that are normally used for the uptake of intermediary metabolites. Noting that the protein:lipid ratio of biomembranes is typically 3:1 to 1:1 and that proteins vary in mass and density (a typical density is 1.37 g/ml) as does their extension, for example, normal to the ca. 4.5 nm lipid bilayer region, the figure attempts to portray a section of a membrane with realistic or typical sizes and amounts of proteins and lipids. Typical protein areas when viewed normal to the membrane are 30%, membranes are rather more ‘mosaic’ than ‘fluid’ and there is some evidence that there might be no genuinely ‘free’ bulk lipids (typical phospholipid masses are 750 Da) in biomembranes that are uninfluenced by proteins. Also shown is a typical drug: atorvastatin (LipitorW) – with a molecular mass of 558.64 Da – for size comparison purposes. If proteins are modelled as
cylinders, a cylinder with a diameter of 3.6 nm and a length of 6 nm has a molecular mass of ca. 50 kDa. Note of course that in a ‘static’ picture we cannot show the dynamics of either phospholipid chains or lipid or protein diffusion.

‘Newly discovered’ metabolites and/or their roles

To illustrate the ‘unfinished’ nature even of Recon2, which concentrates on the metabolites created via enzymes encoded in the human genome, and leaving aside the more exotic metabolites of drugs and foodstuffs and the ‘secondary’ metabolites of microorganisms, there are several examples of interesting ‘new’ (i.e. more or less recently recognised) human metabolites or roles thereof that are worth highlighting, often from studies seeking biomarkers of various diseases – for caveats of biomarker discovery, which is not a topic that we are covering here, and the need for appropriate experimental design. In addition, classes of metabolites not well represented in Recon2 are oxidised molecules such as those caused by nonenzymatic reaction of metabolites with free radicals such as the hydroxyl radical generated by unliganded iron. There is also significant interest in using methods of determining small molecules such as those in the
metabolome (inter alia) for assessing the ‘exposome’, in other words all the potentially polluting agents to which an
individual has been exposed.

Recently discovered effects of metabolites on enzymes 

Another combinatorial problem reflects the fact that in molecular enzymology it is not normally realistic to assess every possible metabolite to determine whether it is an effector (i.e.activator or inhibitor) of the enzyme under study. Typical proteins are highly promiscuous and there is increasing evidence for the comparative promiscuity of metabolites
and pharmaceutical drugs. Certainly the contribution of individual small effects of multiple parameter changes can have substantial effects on the potential flux through an overall pathway, which makes ‘bottom up’ modelling an inexact science. Even merely mimicking the vivo (in Escherichia coli) concentrations of K+, Na+, Mg2+, phosphate, glutamate, sulphate and Cl significantly modulated the activities of several enzymes tested relative to the ‘usual’ assay conditions. Consequently, we need to be alive to the possibility of many (potentially major) interactions of which we are as yet ignorant. One class of example relates to the effects of the very widespread post-translational modification on metabolic
enzyme activities.

A recent and important discovery (Fig. 4) is that a single transcriptome experiment, serving as a surrogate for fluxes through individual steps, provides a huge constraint on possible models, and predicts in a numerically tractable way and
with much improved accuracy the fluxes to exometabolites without the need for such a variable ‘biomass’ term. Other recent and related strategies that exploit modern advances in ‘omics and network biology to limit the search space in constraint-based metabolic modelling.

Fig 4. Workflow for expression-profile-constrained metabolic flux estimation

1. Genome-scale metabolic model with gene-protein-reaction relationships
2. Map absolute gene expression levels to reactions
3. Maximise correlation between absolute gene expression and metabolic flux
4. Predict fluxes to exometabolites
5. Compare predicted with experimental fluxes to exometabolites

Drug Discovery Today

The steps in a workflow that uses constraints based on (i) metabolic network stoichiometry and chemical reaction properties (both encoded in the model) plus, and (ii) absolute (RNA-Seq) transcript expression profiles to enable the
accurate modelling of pathway and exometabolite fluxes. .

Concluding remarks – the role of metabolomics in systems pharmacology

What is becoming increasingly clear, as we recognize that to understand living organisms in health and disease we must treat them as systems, is that we must bring together our knowledge of the topologies and kinetics of metabolic networks with our knowledge of the metabolite concentrations (i.e. metabolomes) and fluxes. Because of the huge constraints imposed on metabolism by reaction stoichiometries, mass conservation and thermodynamics, comparatively few well-chosen ‘omics measurements might be needed to do this reliably (Fig. 4). Indeed, a similar approach exploiting constraints has come to the fore in denovo protein folding and interaction studies.

What this leads us to in drug discovery is the need to develop and exploit a ‘systems pharmacology’ where multiple binding targets are chosen purposely and simultaneously. Along with other measures such as phenotypic screening, and the integrating of the full suite of e-science approaches, one can anticipate considerable improvements in the rate of discovery of safe and effective drugs.

Metabolomics: the apogee of the omics trilogy
Gary J.!Patti, Oscar Yanes and Gary Siuzdak

Metabolites, the chemical entities that are transformed during metabolism, provide a functional readout of cellular biochemistry. With emerging technologies in mass spectrometry, thousands of metabolites can now be
quantitatively measured from minimal amounts of biological material, which has thereby enabled systems-level analyses. By performing global metabolite profiling, also known as untargeted metabolomics, new discoveries linking cellular pathways to biological mechanism are being revealed and are shaping our understanding of cell biology, physiology and medicine.

Metabolites are small molecules that are chemically transformed during metabolism and, as such, they provide a functional readout of cellular state. Unlike genes and proteins, the functions of which are subject to epigenetic regulation and posttranslational modifications, respectively, metabolites serve as direct signatures of biochemical activity and are therefore easier to correlate with phenotype. In this context, metabolite profiling, or metabolomics, has become a powerful approach that has been widely adopted for clinical diagnostics.

The field of metabolomics has made remarkable progress within the past decade and has implemented new tools that have offered mechanistic insights by allowing for the correlation of biochemical changes with phenotype.

In this Innovation article, we first define and differentiate between the targeted and untargeted approaches to metabolomics. We then highlight the value of untargeted metabolomics in particular and outline a guide to performing such studies. Finally, we describe selected applications of un targeted metabolomics and discuss their potential in cell biology.

• metabolites serve as direct signatures of biochemical activity

1. In some instances, it may be of interest to examine a defined set of metabolites by using a targeted approach.
2. In other cases, an untargeted or global approach may be taken in which as many metabolites as possible are measured and compared between samples without bias.
3. Ultimately, the number and chemical composition of metabolites to be studied is a defining attribute of any metabolomic experiment and shapes experimental design with respect to sample preparation and choice of instrumentation.

The targeted and untargeted workflow for LC/MS-based metabolomics.

a | In the triple quadrupole (QqQ)-based targeted metabolomic workflow, standard compounds for the metabolites of interest are first used to set up selected reaction monitoring methods. Here, optimal instrument voltages are determined and response curves are generated for absolute quantification. After the targeted methods have been established
on the basis of standard metabolites, metabolites are extracted from tissues, biofluids or cell cultures and analysed. The data output provides quantification only of those metabolites for which standard methods have been built.

b | In the untargeted metabolomic workflow, metabolites are first isolated from biological samples and subsequently analysed by liquid chromatography followed by mass spectrometry (LC/MS). After data acquisition, the results are processed by using bioinformatic software such as XCMS to perform nonlinear retention time alignment and identify peaks that are changing between the groups of samples measured. The m/z value s for the peaks of interest are searched in metabolite databases to obtain putative identifications. Putative identifications are then confirmed
by comparing tandem mass spectrometry (MS/MS) data and retention time data to that of standard compounds. The untargeted workflow is global in scope and outputs data related to comprehensive cellular metabolism.

Metabolic Biomarker and Kinase Drug Target Discovery in Cancer Using Stable Isotope-Based Dynamic Metabolic Profiling (SIDMAP)

László G. Boros1*, Daniel J. Brackett2 and George G. Harrigan3
1UCLA School of Medicine, Harbor-UCLA Research and Education Institute, Torrance, CA. 2Department of Surgery, University of Oklahoma Health Sciences Center & VA Medical Center, Oklahoma City, OK, 3Global High Throughput
Screening (HTS), Pharmacia Corporation, Chesterfield, MO.
Current Cancer Drug Targets, 2003, 3, 447-455.

Tumor cells respond to growth signals by the activation of protein kinases, altered gene expression and significant modifications in substrate flow and redistribution among biosynthetic pathways. This results in a proliferating phenotype
with altered cellular function. These transformed cells exhibit unique anabolic characteristics, which includes increased and preferential utilization of glucose through the non-oxidative steps of the pentose cycle for nucleic acid synthesis but limited denovo fatty acid synthesis and TCA cycle glucose oxidation. This primarily nonoxidative anabolic profile reflects an undifferentiated highly proliferative aneuploid cell phenotype and serves as a reliable metabolic biomarker to determine cell proliferation rate and the level of cell transformation/differentiation in response to drug treatment. Novel drugs effective in particular cancers exert their anti-proliferative effects by inducing significant reversions of a few specific non-oxidative anabolic pathways. Here we present evidence that cell transformation of various mechanisms is sustained by a unique
disproportional substrate distribution between the two branches of the pentose cycle for nucleic acid synthesis, glycolysis and the TCA cycle for fatty acid synthesis and glucose oxidation. This can be demonstrated by the broad labeling and unique specificity of [1,2-13C2]glucose to trace a large number of metabolites in the metabolome. Stable isotope-based dynamic metabolic profiles (SIDMAP) serve the drug discovery process by providing a powerful new tool that integrates the metabolome into a functional genomics approach to developing new drugs. It can be used in screening kinases and their metabolic targets, which can therefore be more efficiently characterized, speeding up and improving drug testing, approval and labeling processes by saving trial and error type study costs in drug testing.

Navigating the HumanMetabolome for Biomarker Identification and Design of Pharmaceutical Molecules

Irene Kouskoumvekaki and Gianni Panagiotou
Department of Systems Biology, Center for Biological Sequence Analysis, Building 208, Technical University of Denmark, Lyngby, Denmark
Hindawi Publishing Corporation  Journal of Biomedicine and Biotechnology 2011, Article ID 525497, 19 pages
http://dx.doi.org:/10.1155/2011/525497

Metabolomics is a rapidly evolving discipline that involves the systematic study of endogenous small molecules that characterize the metabolic pathways of biological systems. The study of metabolism at a global level has the potential to contribute significantly to biomedical research, clinical medical practice, as well as drug discovery. In this paper, we present the most up-to-date metabolite and metabolic pathway resources, and we summarize the statistical, and machine-learning tools used for the analysis of data from clinical metabolomics.

Through specific applications on cancer, diabetes, neurological and other diseases, we demonstrate how these tools can facilitate diagnosis and identification of potential biomarkers for use within disease diagnosis. Additionally, we
discuss the increasing importance of the integration of metabolomics data in drug discovery. On a case-study based on the Human Metabolome Database (HMDB) and the Chinese Natural Product Database (CNPD), we demonstrate the close relatedness of the two data sets of compounds, and we further illustrate how structural similarity with human metabolites could assist in the design of novel pharmaceuticals and the elucidation of the molecular mechanisms of medicinal plants.

Metabolites are the byproducts of metabolism, which is itself the process of converting food energy to mechanical energy
or heat. Experts believe there are at least 3,000 metabolites that are essential for normal growth and development (primary metabolites) and thousands more unidentified (around 20,000, compared to an estimated 30,000 genes and 100,000 proteins) that are not essential for growth and development (secondary metabolites) but could represent prognostic, diagnostic, and surrogate markers for a disease state and a deeper understanding of mechanisms of disease.

Metabolomics, the study of metabolism at the global level, has the potential to contribute significantly to biomedical
research, and ultimately to clinical medical practice. It is a close counterpart to the genome, the transcriptome and the proteome. Metabolomics, genomics, proteomics, and other “-omics” grew out of the Human Genome Project, a massive research effort that began in the mid-1990s and culminated in 2003 with a complete mapping of all the genes in the human body. When discussing the clinical advantages of metabolomics, scientists point to the “real-world” assessment
of patient physiology that the metabolome provides since it can be regarded as the end-point of the “-omics” cascade. Other functional genomics technologies do not necessarily predict drug effects, toxicological response, or disease states at the phenotype but merely indicate the potential cause for phenotypical response. Metabolomics can bridge this information gap. The identification and measurement of metabolite profile dynamics of host changes provides the closest link to the various phenotypic responses. Thus it is clear that the global mapping of metabolic signatures pre- and postdrug treatment is a promising approach to identify possible functional relationships between medication and medical phenotype.

Human Metabolome Database (HMDB). Focusing on quantitative, analytic, or molecular scale information about
metabolites, the enzymes and transporters associated with them, as well as disease related properties the HMDB represents the most complete bioinformatics and chemoinformatics medical information database. It contains records for
thousands of endogenous metabolites identified by literature surveys (PubMed, OMIM, OMMBID, text books), data
mining (KEGG, Metlin, BioCyc) or experimental analyses performed on urine, blood, and cerebrospinal fluid samples.
The annotation effort is aided by chemical parameter calculators and protein annotation tools originally developed for
DrugBank.

A key feature that distinguishes the HMDB from other metabolic resources is its extensive support for higher level database searching and selecting functions. More than 175 hand-drawn-zoomable, fully hyperlinked human
metabolic pathway maps can be found in HMDB and all these maps are quite specific to human metabolism and
explicitly show the subcellular compartments where specific reactions are known to take place. As an equivalent to
BLAST the HMDB contains a structure similarity search tool for chemical structures and users may sketch or
paste a SMILES string of a query compound into the Chem-Query window. Submitting the query launches a
structure similarity search tool that looks for common substructures from the query compound that match the
HMDB’s metabolite database. The wealth of information and especially the extensive linkage to metabolic diseases
to normal and abnormal metabolite concentration ranges, to mutation/SNP data and to the genes, enzymes, reactions
and pathways associated with many diseases of interest makes the HMDB one the most valuable tool in the hands
of clinical chemists, nutritionists, physicians and medical geneticists.

Metabolomics in Drug Discovery and Polypharmacology Studies

Drug molecules generally act on specific targets at the cellular level, and upon binding to the receptors, they exert
a desirable alteration of the cellular activities, regarded as the pharmaceutical effect. Current drug discovery depends
largely on ransom screening, either high-throughput screening (HTS) in vitro, or virtual screening (VS) in silico. Because
the number of available compounds is huge, several druglikeness filters are proposed to reduce the number of compounds that need to be evaluated. The ability to effectively predict if a chemical compound is “drug-like” or “nondruglike” is, thus, a valuable tool in the design, optimization, and selection of drug candidates for development. Druglikeness is a general descriptor of the potential of a small molecule to become a drug. It is not a unified descriptor
but a global property of a compound processing many specific characteristics such as good solubility, membrane
permeability, half-life, and having a pharmacophore pattern to interact specifically with a target protein. These
characteristics can be reflected as molecular descriptors such as molecular weight, log P, the number of hydrogen bond
donors, the number of hydrogen-bond acceptors, the number of rotatable bonds, the number of rigid bonds, the
number of rings in a molecule, and so forth.

Metabolomics for the Study of Polypharmacology of Natural Compounds

Internationally, there is a growing and sustained interest from both pharmaceutical companies and public in medicine
from natural sources. For the public, natural medicine represent a holistic approach to disease treatment, with
potentially less side effects than conventional medicine. For the pharmaceutical companies, bioactive natural products
constitute attractive drug leads, as they have been optimized in a long-term natural selection process for optimal interaction with biomolecules. To promote the ecological survival of plants, structures of secondary products have evolved to interact with molecular targets affecting the cells, tissues and physiological functions in competing microorganisms,
plants, and animals. In this, respect, some plant secondary products may exert their action by resembling endogenous
metabolites, ligands, hormones, signal transduction molecules, or neurotransmitters and thus have beneficial
effects on humans.

Future Perspectives

Metabolomics, the study of metabolism at the global level, is moving to exciting directions.With the development ofmore
sensitive and advanced instrumentation and computational tools for data interpretation in the physiological context,
metabolomics have the potential to impact our understanding of molecular mechanisms of diseases. A state-of-theart
metabolomics study requires knowledge in many areas and especially at the interface of chemistry, biology, and
computer science. High-quality samples, improvements in automated metabolite identification, complete coverage of
the human metabolome, establishment of spectral databases of metabolites and associated biochemical identities, innovative experimental designs to best address a hypothesis, as well as novel computational tools to handle metabolomics data are critical hurdles that must be overcome to drive the inclusion of metabolomics in all steps of drug discovery and drug development. The examples presented above demonstrated that metabolite profiles reflect both environmental and genetic influences in patients and reveal new links between metabolites and diseases providing needed prognostic,diagnostic, and surrogate biomarkers. The integration of these signatures with other omic technologies is of utmost importance to characterize the entire spectrum of malignant phenotype.