This chapter covers strategic and practical aspects related to optimal ways in which biomarkers for translational and personalized medicine can be applied to innovate pharmaceutical drug development, and contribute to improved health and disease management.

Biomarkers have been around since the beginning of medicine when the colour of skin, various characteristics of urine (exemplified by the diagnostic “urine wheel” published in 1506 by Ullrich Pinder, in his book Epiphanie Medicorum¹ ), and other qualitative assessments were interpreted as biological markers of a person’s well‐being. For a long time, phenotypic analyses combined with a patient’s self‐assessment were the only tools for diagnosis of disease and monitoring of treatment effects. Recent breakthroughs in molecular technologies to identify, understand and measure biomarkers have strongly increased the possibilities towards a person‐specific assessment of disease. These include accurate prediction of a person’s risk to develop a specific disease, early detection of a prevalent disease, prediction of disease progression, and prediction and monitoring of the effects of disease treatment, all in a personalized manner.

Biomarkers can be diverse. Ten years ago a useful definition of a biomarker was drafted, being “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention”.²  There are several important aspects in this definition, both in terms of what is described and what is missing. First, a biomarker can be an indicator of a normal process, of a derailed process related to disease or of the effects of a certain treatment thereon. Although this covers many of the applications, biomarker scientists argue it does not describe biomarkers that indicate disease risk, e.g. through genetic predisposition or brought about by a certain lifestyle. Secondly, the biomarker is a characteristic, meaning it can have multiple identities ranging from a single protein in serum to a complex three‐dimensionally reconstructed image of the brain. This has caused several discussions in the field; indeed, would an established biochemical assay that has been operational long before the current biomarker hype, such as estadiol analysis, qualify as a biomarker? If so, how about a mechanical read‐out such as a pressure meter in a pen, used by anxiety patients filling in a questionnaire? How about the questionnaire itself? Thirdly, the biomarker is to be objectively measured and evaluated, implying the biomarker assay read‐out is trustable and actionable. This leaves room to decide exactly how objective a biomarker should be measured before enabling a clinical decision, fueling discussions on fit‐for‐purpose robustness of the assay. Despite these alternative views, the stated definition is still a useful one and used by many to focus their attentions to the output defined.

Because of their potential in clinical applications, biomarkers have received much interest in the biomedical field. Their main applications seem to reside in two areas. In translational medicine, knowledge from preclinical models is translated to clinical practice and back using biomarkers that can reflect various aspects of a biological system including molecular pathways, functional cell–cell interactions and tissue metabolism. Such studies are expected to greatly increase the molecular knowledge of the mechanisms of human disease and pathophysiology, leading to a better diagnosis and more effective clinical treatment. In personalized medicine, biomarkers are used to profile patients and to define which treatment should be given to which patient at what time and at what dose. Such stratification biomarkers are expected to strongly increase the chance of a successful clinical treatment by selecting patients that are most likely to respond to a drug and/or to deselect patients that are predicted to exhibit adverse effects.

The availability of the human genome sequence in the late 1900s prompted many to believe that by 2010 personalized medicine would be fully implemented as each person would have his/her genome on a chip to enable a physician to determine the best personalized care. Former president of the USA Bill Clinton phrased it in his 1998 State of the Union Address as: “Gene chips will offer a road map for prevention of illness through a lifetime”. There are many shining examples where hard work has indeed resulted in good clinical utility of biomarkers, but there still is a long way to go, as discussed in this chapter.

Biomarkers have become part of our daily lives as illustrated by advertisement of the positive effects of nutrition based on biomarkers (e.g. cholesterol lowering), by media‐supported general education about the molecular processes in a human body and how biomarkers represent those processes, by availability and acceptance of biomarker‐based “health checks” that can be performed through dedicated providers or even main‐street pharmacies, by smartphone apps that provide a health check based on biomarker data, and so on. This all leads to more aware and vocal patients who debate with their physician about their best treatment, rather than “following doctor’s orders”.

Regarding industrial implementation, biomarkers in the pharmaceutical drug development field have been leading the way, as they matured from explorative pharmacological parameters to essential tools to characterize a patient in molecular detail and to monitor drug action after dosing. In development of neutriceuticals (functional ingredients of food) biomarkers can have a similar role and potentially similar biomarkers representing biological mechanisms or metabolic physiological states can be used. Also, cosmetics can be an interesting biomarker application area, whereby the biomarker read‐outs can demonstrate absence of side‐effects of the cosmetics. Interestingly, biomarkers are receiving increasing interest to quantify health. Health is described to be “not merely the absence of disease but the ability to adapt to one’s environment”, also called resilience.³  Health biomarkers thus indicate the risk of an individual to develop a disease and can be key drivers of prevention strategies, including timely correction by lifestyle change, neutriceuticals or pharmaceuticals.

Despite these positive developments, it was anticipated that progress in translational and personalized medicine would be more advanced than it is today. The discovery of a biomarker and its maturation to a clinically usable test has been shown to require a thorough and long‐lasting research and development process.

In this chapter we will mainly focus on biomarkers in pharmaceutical drug development, as lessons learned there can be applied to biomarkers in other application areas. After outlining how biomarkers do play a role in decision making during development of drugs, and more specifically their role in translational and personalized medicine, we will review trends, challenges and opportunities related to biomarkers in biomedical science.

Before discussing the role of biomarkers in pharmaceutical drug development, translational medicine and personalized medicine, we would first like to list useful definitions of biomarkers and their utilities in this field that will guide the thought process.⁴ 

• Biomarker: A characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.² 

• Disease biomarkers: Biomarkers that are correlated with the disease where correlation is established via rigorous biological and clinical validation. Disease biomarkers are not necessarily causally associated with the mechanism of the disease. Correlation to important phenotypes of the disease, relationships to its initiation, propagation, regression or relapses, however, must be established. Disease biomarkers can serve as diagnostic biomarkers (distinguishing patients from nonpatients), as prognostic biomarkers (identifying “rapid vs. slow progressing” patients) or as disease‐classification biomarkers (elucidating molecular mechanisms of the observed pathophysiology). All three functions are crucial parameters to drive the selection of subjects in clinical studies.⁵ 

• Target Engagement Biomarker: Biomarkers that represent the direct interaction of the drug (small molecule or biological) with the molecular target. These are highly important to guide drug exposure as they reflect distribution of the drug to the specific location of target, the residency time of the drug on the target and the extent of the drug target modulation by the bound drug.

• Pharmacokinetic Biomarkers: Biomarkers that represent the level of the pharmaceutical drug in circulating body fluids and/or at the site of action, and that are important to calculate the dose needed to induce a certain pharmacological response.

• Pharmacodynamic Biomarkers: Biomarkers that represent the functional outcome of the interaction of a drug with its target (also called pharmacological biomarkers). These biomarkers can have various identities, can be analyzed by a variety of methodologies (including enzymology, omics, imaging), but generally represent a read‐out of complex biology. Pharmacodynamic biomarkers are specifically used to rationalize clinical therapeutic efficacy and adverse effects, typically measured as a multiparameter panel of biomarkers representing distinct functional events.

• Predictive Biomarker: Biomarkers that are used for the selection of patients for clinical studies. These biomarkers serve to predict which patients are likely to respond to a particular treatment or drug's specific mechanism of action, or potentially predict those patients who may experience adverse effects.

• Validated Biomarkers: Biomarkers that are measured in an analytical test system with well‐established performance characteristics, and with established scientific framework or body of evidence that elucidates the physiologic, pharmacologic, toxicologic, or clinical significance of the test results.⁶ 

• Surrogate Endpoint: A biomarker that is intended to substitute for a clinical endpoint. A surrogate endpoint is expected to predict clinical benefit (or harm or lack of benefit or harm) based on epidemiologic, therapeutic, pathophysiologic, or other scientific evidence.² 

• Therapeutic Biomarker: A biomarker that indicates the effect of a therapeutic intervention and can be used to assess its efficacy and/or safety liability.¹⁰ 

The pharmaceutical drug‐development process (Figure 1.1) is a multistep process that on average takes 14 years from initiation of research to marketing of the new medicine.

It starts with the discovery of a drug target, the molecular target that the drug will act upon. In this target discovery phase researchers investigate how a particular disease is caused and what factors play a key role. The inhibition or stimulation of those key factors will be the basis for the new pharmaceutical drug used to treat the selected disease. For example, postmenopausal complaints are caused by a decrease in endogenously produced estrogens and the objective is to find estrogen‐like compounds that can supplement the natural pool of estrogens. The drug target in this case is the estrogen receptor, which is a nuclear protein present in cells of specific tissues and acts as an estrogen‐activated transcription factor with specific effects on each specific tissue. Indeed, the activated estrogen receptor in osteoblasts mediates the synthesis of new bone, whereas the estrogen receptor in breast epithelial cells is a key player in cell growth.

The next step is the identification of compounds that have the desired effect on the drug target, for example by inhibiting or stimulating its activity. This discovery is done in the lead discovery phase, during which large numbers of chemical or biological compounds are tested for the desired effects in biochemical and cellular assays. Often mechanistic biomarkers or derivatives thereof are used as read‐outs of the screening assays. In the estrogen receptor example, a screening assay to identify compounds with agonistic or antagonistic estrogenic activity may comprise of a cell line expressing the receptor and containing an estrogen‐receptor sensitive luciferase reporter module.

Following selection of the most promising hits and limited optimization, the lead optimization phase starts where systematically up to a thousand variants of the original positive substances are synthesized and tested in various tests. A stringent selection process aims to select those compounds that display improved efficacy, specificity, safety, bioavailability and/or production efficiency (depending on the objectives of the project). Assays used during lead optimization include biochemical and cellular test systems, followed by in vivo assays to assess bioavailability, pharmacological and toxicological effects of the drug. Typically one or two of the best compounds are nominated to progress from research to development.

In preclinical development the substance is first investigated in animal models to test whether it is sufficiently bioavailable and safe. After successfully passing this phase, similar studies are performed in the phase 1 clinical trials, during which under strictly controlled conditions the compound is tested in human subjects. Typically, healthy human subjects participate in such trials; the exception being oncology trials whereby drugs are often tested directly in small numbers of patients. Subsequently, the compound is tested in patients, which is the first time the drug developer will determine whether the originally chosen approach of affecting the drug target has a positive effect on treatment of the disease. This occurs in a phase 2 clinical trial in which a relatively small group of patients are tested. A positive outcome of this trial, with an acceptable level of side effects, is very important as it is then proved that the approach chosen to treat the disease works, also known as the clinical proof of concept. After this milestone, the clinical phase 3 starts in which the effect of the substance is tested on large numbers of patients. Such a study can be very substantial. For instance, a phase 3 trial of testing estrogenic compounds in osteoporosis involves administration of the candidate drug in thousands of postmenopausal women per dose group for three years while recording how often a participant breaks a hip, a reduction of which is the currently accepted clinical endpoint of efficacy.⁷ 

The final stage of the drug‐development process involves scaling up production, obtaining regulatory approvals, designing the drug label and packaging, and preparing the market launch. In addition, after introducing the drug on the market, often additional scientific studies are done to further elucidate the mechanism of action or to investigate whether the drug can be applied in other therapeutic indications.

Despite the strong underlying science and process‐minded workflow of drug discovery and development projects, this process is highly inefficient. It takes on average 14 years to bring a new drug to the market starting from a new drug‐discovery project, at the cost of 1.6 billion US dollars.⁹  The main cause of this resides in the high attrition rate of 90% during clinical development, meaning that only one in ten projects being pursued through clinical testing in patients is successful^10,11  An analysis of various therapeutic areas has illustrated that there are differences in success rate in different therapeutic areas; drugs treating CNS, oncology and women’s health are the most difficult to develop.^10,11  In particular, the proof‐of‐concept phase contributes to the high attrition because of a lack of efficacy and/or unacceptable safety liabilities. This leads one to surmise that preclinical studies in pharmaceutical research are insufficient to predict drug action in patients, resulting in a strong need for translational biomarkers to bridge research and development.

Although the development of a drug is described above as a linear process, in reality it is often an intertwined process with many forward and backward translation loops. Indeed, to better define the objective of target discovery, a strong interaction is needed between the molecular biologists selecting the drug target and the clinical experts that define the patient needs. Also, a strong interaction is needed during lead optimization between pharmacologists and early clinical developers to ensure smooth transition of the compound from research to development and timely preparation of the clinical assays needed. Such translational medicine is supported by high‐quality biomarkers that enable monitoring of drug action at all critical stages of drug development (Figure 1.2).

Most, if not all, pharmaceutical companies developed generic biomarker strategies to support their drug‐development projects through a rational and consistent application of key decision‐making biomarkers. Previously we published a question‐based drug development biomarker strategy that is based on a set of translational questions.¹²  In this strategy, answers are provided by translational biomarkers, enabling data‐driven decisions throughout the development of a drug. The questions in this question‐based drug development approach include:

• Does the compound get to the site of action?

• Does the compound cause its intended pharmacological and functional effects?

• Does the compound have beneficial effects on disease or clinical pathophysiology?

• What is the therapeutic window?

• How do sources of variability in drug response in target population affect efficacy and safety?

The emphasis on each question is different for each drug‐development project and varies depending on therapeutic area, mechanism and drug target. In the example of the estrogen receptor and postmenopausal osteoporosis, biomarkers are mostly needed to reflect the tissue‐specific modulation of the estrogenic pathway, particularly the pharmacological effects of the estrogen‐like compounds on bone, and the absence of undesired extrogenic effects on breast and endometrium.¹³  In some cases, a biomarker strategy and/or selected biomarkers can be used for other projects whereas similar approach is being followed, for instance when using the estrogen receptor α biomarkers for characterization of estrogen receptor β modulators.¹⁴ 

In all cases, however, a target engagement biomarker is essential, indicating the physical or mechanistic engagement of the drug target by the drug compound. Together with read‐outs of clinical efficacy this will drive early decisions whether or not to proceed with a clinical development project, following the decision tree outlined in Figure 1.3. When the compound does not engage the drug target sufficiently to levels determined in preceding research, it makes no sense to test patients for drug efficacy. One needs to identify a more potent compound or improved administration route, but the original concept that relates the drug target to the disease may still be correct. If, however, the compound shows sufficient engagement of the drug target but there is insufficient effect on the disease, then the concept was shown to be incorrect and a (data‐driven) decision to abandon the approach should be taken. In the most positive scenario, sufficient target engagement is followed by a positive effect on the disease, which in combination delivers the highly desired clinical Proof‐of‐Concept. This important milestone will be the basis to decide for extensive clinical testing in full development.

The purpose of this biomarker‐driven decision making is to decide as early in the process as possible on the best drug candidates and select the most successful approaches. This rational biomarker‐driven decision making could shorten the development time per compound, but will surely improve the overall efficiency of the pipeline and reduce attrition later, providing an overall reduction in costs and time. In particular, late‐stage failures in phase 3 are very costly due to the clinical trial itself but also due to the preparations to produce the new pharmaceutical ingredient on the large scale once the clinical tests were positively concluded. This strategy is also designated as “more shots on goal” and “stop early, stop cheap’.

For each project, it is of high importance to define early in the project the exact quantitative criteria that are to be used to evaluate the drug response using the biomarker data. Often it is not yet known what the thresholds of the biomarker are, for instance whether a change of 20% or 30% is needed to define a positive response. In some cases, this is particularly hard as the key disease‐specific biomarkers can be below detection level in samples from healthy subjects. If so, one cannot define a priori to what extent a disease‐specific biomarker should decrease following treatment to healthy levels. Clinical biomarker validation efforts should be strongly focused on this aspect, and the information should be translated back to preclinical studies where similar thresholds should be regarded. In particular, the relationship between drug exposure (pharmacokinetics, PK) and the changing levels of the biomarker (pharmacodynamics, PD) enable a forward translation of preclinical findings through modeling and simulation. Such PK/PD model can quantitatively predict the change to be expected in clinical testing and can lead to decisions whether the window between drug efficacy and safety is large enough for clinical success. In a truly translational medicine workflow, such quantitative analysis of key biomarkers should start at drug target validation, continue at lead optimization, and be extended to clinical trials.

The majority of biomarker use in drug development is in the early phases for internal decision making on which drug to progress. Once pivotal trials are being run, typically phase 2b and phase 3 in late‐stage development, biomarkers have a less‐explorative role but are used as key decision points. Pivotal trials are used for registration of the drug for a defined indication, and the biomarkers that aid in building the filing dossier should be widely accepted by the scientific clinical communities and regulatory bodies. In some cases, biomarkers from research do progress to late clinical drug development as surrogate endpoints or companion diagnostics.

Surrogate endpoints are those biomarkers that substitute a clinical endpoint with the objective to predict the outcome of a clinical intervention where the primary endpoint is undesired (for instance mortality rate) or rare, requiring very large numbers of patients in the trial.¹⁵  An example of the latter is the aforementioned phase 3 clinical trial in osteoporosis drug development. A surrogate endpoint that would replace such time‐consuming clinical assessment would have added value, and a combination of noninvasive computer tomography imaging of bone structure combined with an algorithm to calculate bone strength based on 3D images¹⁶  is a promising approach in this respect. The maturation of a biomarker to become a surrogate endpoint requires an extensive and lengthy clinical validation process that involves multiple clinical centers, intervention trials, key stakeholders and requires an intensive interaction with regulatory agencies such as FDA, EMA. An example of such surrogate endpoint biomarkers is cholesterol, that was used for registration of simvastatin.¹⁷ 

Companion diagnostics have received much attention in recent years, due to their intrinsic potential to enable personalized medicine. As discussed below, currently there are over one hundred FDA‐approved drugs that contain pharmacogenomic information in their labels, and where the outcome of the biomarker test determines whether people should be prescribed this drug or not. Many of these genomic tests relate to metabolism, related to variants in the Cyp450 gene family, and the majority of applications reside in oncology and psychiatry.¹⁸  To a lesser extent, but increasingly, companion diagnostics are being developed for other disease areas and much is to be expected for the years to come.

Lastly, biomarkers are being applied to the very last phase of drug development, the postmarket introduction phase 4, when the drug is being used in a certain therapeutic indication and additional studies are being performed to either learn more of its mechanism of action or to investigate potential applications in other disease areas. Here, the use of biomarkers is again of a more explorative nature, comparable to the early drug‐development phases.

The different applications of biomarkers to support the pharmaceutical drug‐development process as described above already indicate the variable requirement of the biomarker assay robustness. When still in early research phases, such as the validation of a drug target in pharmacological studies, the biomarker merely serves to indicate functional effects and the corresponding assay does not necessarily have to be very robust. The test can yield qualitative data, e.g. condition 1 versus condition 2, although quantitative data is always preferred. In contrast, if the biomarker read‐out is used to demonstrate clinical proof‐of‐principle or proof‐of‐concept, with decision‐making implementations for the drug‐development program, the assay needs to be robust, reproducible and quantitative. Importantly, the biomarker should have a solid body of evidence indicating its value as the intended read‐out, and be accepted by the scientific clinical community. Thirdly, if the biomarker progresses to a surrogate endpoint or companion diagnostic, a rigorous process to generate a diagnostic test has to be followed. In this phase the focus is on thorough analytical validation and production of the biomarker test with a positive predicted health economic value. Clinical utility should be similar and acceptance of the biomarker test by clinical key opinion leaders and regulatory bodies is crucial. Useful guidelines to facilitate the assay development and support the definition of the fit‐for‐purpose robustness have been published.¹⁹  In general, a robust quantitative biomarker assay is always preferred, as this facilitates translation of findings across different drug‐development phases and data sharing among laboratories.

The success rate of novel medical entities that are submitted for registration by the regulatory agencies and follow successful marketing has been stagnating for the past decade. Failure in efficacy and safety continue to be the prime hurdles and cause of failure and attritions in drug‐discovery process,^20,21  particularly in phase 2 and phase 3 in the pharmaceutical pipelines.^22,23 

Translational medicine is an emerging paradigm within the pharmaceutical industry R&D organizations aimed to improve the predictability and success of drug discovery and development.²⁴  Translational Medicine is the integrated application of innovative pharmacology tools, biomarkers, clinical methods, clinical technologies and study designs to improve confidence in human drug targets, increase confidence in drug candidates, understand the therapeutic index in humans, enhance cost‐effective decision making in exploratory development, all with the objective to increase success in clinical phase 2.

The key objective of translational medicine in the pharmaceutical research is to improve the success rate of compounds and biologics identified in the discovery phase and chosen for clinical development. Towards this goal, translational medicine encompasses many different activities along the drug‐discovery pipeline enumerated in Section 1.3.1. These include:

• Target validation to establish that the molecular target selected for further development contributes significantly to a human disease and that manipulation of the target could provide desired benefits while minimizing potential adverse effect. The key goal of this activity is to establish the “therapeutic window” of safety and tolerability.

• Better understanding of the Pharmacokinetics (PK) and Pharmacodynamic (PD) relationships of the drug with the molecular target. Understanding the pharmacological nature of compound interaction with the target (e.g. target occupancy and target engagement), extent of modulation needed to achieve desired therapeutic efficacy while reducing any “off target” effects.

• Selection of preclinical model systems that provide insights into human disease. Such systems include in vitro biochemical systems, cell‐based assays, ex vivo organ systems and integrated in vivo animal models. A key limitation in this aspect is the lack of sufficient representative cellular and animal models that approximate the human disease and their pathophysiology.

• Patient selection/stratification based on genetic variations for rational clinical trial designs. In recent years, substantial efforts have been made to carefully select patients that are likely to respond to a particular treatment regime. Cancer drug discovery and development is the prominent example of this paradigm shift. Modern molecular oncology seeks the discovery and development of drugs that specifically address the mechanism of the oncogenic transformation in particular cancer types so that the drugs are tailored towards them. In a very recent example FDA approved a new drug for cystic fibrosis on the basis of an expedited (3‐month) review and smaller clinical studies of a compound VX‐770 (Kalydeco) that targeted a rare mutation of CFTR gene (G551D).²⁵  Proper application of stratification biomarkers in development of preclinical models will ensure smooth translation of findings to clinical development.

Animal models have been at the core of the drug‐discovery process spanning nearly all phases of preclinical and clinical R&D phases. Rodent models have significantly contributed to our molecular and cellular understanding of various diseases, for instance various types of cancer, metabolic, immunological and neurological diseases. Genetically engineered animal models have enabled investigators to both observe and manipulate complex disease processes in a manner impossible to perform in patients²⁶  and have led to successful clinical translation of biological insights, for instance in the treatment of hitherto rare yet fatal acute promyelocytic leukemia (APL) that is now very effectively cured and managed in patients.^27,28 

However, animal models also have their own limitations due to species‐specific genetic differences with regard to humans resulting in inaccurate and at times inadequate recapitulation of human disease pathophysiology. Mounting evidence also suggests that there are limitations on how closely some of the established models mimic human physiology leading to failure in accurately predicting the effects of drugs in human subjects, for instance in neurodegenerative and immunological diseases.^29,30 

In the context of immunology there are key molecular differences between mouse and human, which should be dealt with cautiously when using mouse as translational model for studying specific aspects of human immunology. For instance, important differences are seen in expression of key cellular proteins related to inflammation, for instance S100A12 (ENRAGE), which binds to RAGE (receptor of advanced glycation endproducts), and whose ligation has been implicated in various inflammatory‐related diseases.³¹  However, this mechanism cannot be studied in mouse as there is no evidence of functional murine S100A12 gene expression.³²  Similarly, a wide repertoire of chemokines have been identified in humans but not in mice, including IL‐8 (CXCL8), neutrophil‐activating peptide‐2 (CXCL7), IFN‐inducible T cell α‐chemoattractant (CXCL11), monocyte chemoattractant protein (MCP)‐4 (CCL13), HCC‐1 (CCL14), hemofiltrate CC chemokines‐2 (CCL15), pulmonary and activation‐regulated chemokine (CCL18), myeloid progenitor inhibitory factor‐1 (CCL23), and eotaxin‐2/3 (CCL24/CCL26). Conversely, CCL6, CCL9, lungkine (CXCL15), and MCP‐5 (CCL12) have been identified in mice but not humans.

A proper evaluation and definition of translational biomarkers could mitigate these limitations leading to improved translation of results from preclinical models to human. This approach requires components such as:

• Selection of the right species that allows monitoring biomarkers that can be monitored in preclinical models, healthy subjects and patients. For instance, nonhuman primates (NHP) to study key human diseases (metabolic, neurodegenerative and immunological) have for this reason gathered momentum in recent years.³³ 

• Biomarkers that enable translational PK/PD relationship in preclinical models and in human disease.

• Conservation of the (patho)physiological process(es) studied in the animal with the human disease. whereby the functional role of biomarkers converge at spatial, temporal and functional levels across the experimental models and human studies. This principle in turn necessitates looking at cellular processes and their perturbations at systems levels. Humanized rodent models such as primary tumor transplant models may form such a hybrid translational test system.³⁴ 

Rational applications of these principles have shown encouraging results in contemporary drug discovery. For instance, in strokes whereby studies in nonhuman primate (NHP) models have led to identification of PSD95 inhibitors that show promising results for stroke treatment, thus opening a new vista for future clinical validation^35,36  (Figure 1.4A). Similarly, clinical trials to treat a rare form of cancer, i.e. pancreatic neuroendocrine tumors (PNET) have been undertaken with encouraging phase III results, based on elaborate studies performed in mouse models^37,38  (Figure 1.4B). In metabolic disease space recently an engineered FGF21‐mimetic monoclonal antibody “mimAb1” was tested in obese cynomolgus monkeys (NHP model) leading to decrease in body weight and body mass index (BMI).³⁹  Furthermore, mimAb1 treatment in NHP model led to decreases in fasting and fed plasma insulin levels, as well as a reduction in plasma triglyceride and glucose levels. While these results need to be replicated in human cohorts they nevertheless illuminate a beneficial therapeutic approach towards treating patients with diet‐induced obesity and diabetes. In yet another study, a three‐part therapy approach to cure type 1 diabetes was tested in a type 1 diabetes mouse model, which holds promises for multifaceted approach for curing this autoimmune disease.⁴⁰ 

Until recently the approach towards medicine (and healthcare) has been largely population‐driven and reactive, whereby emphasis was placed on treating disease symptoms rather than on preventing disease. Predicated on this model, the drug‐discovery process was also focused on developing drugs for masses that would have high return on R&D investments. The majority of the current drugs are approved and developed on the basis of their performance in a large population of people and each drug is prescribed to all patients with a certain diagnosis.

However, it is clear that this “blockbuster” model does not apply to all therapeutic areas. Indeed, hypertension drugs as ACE inhibitors are only efficacious in 10–30% of patients, heart failure drugs as beta blockers in only 15–25%, antidepressants in 20–50% of patients and so on.⁴¹  With the advent of various molecular profiling sciences it has become feasible to investigate this in great detail at the genetic, transcriptomic, epigenetic, proteomic, metabolomic and other complementary “Omic” levels. The resultant knowledge and insights have led to the appreciation that most of the diseases are extremely heterogeneous and the “one‐size fits all” approach to treatment is ineffective and unsustainable.⁴²  In the postgenomics era a new approach to medicine called “Personalized Medicine” is taking shape that is informed and driven by each person’s unique clinical, genetic, genomic, and environmental information.⁴³  Personalized medicine is a paradigm shift in healthcare. Its success depends on multidisciplinary biomedicine, adaptive clinical practices and integrative technologies to leverage our molecular understanding of disease towards optimized preventive health care strategies.^44,45  This new healthcare model predicates a central role of the emerging and empowered patient who actively participates in his health and disease management,⁴⁶  as part of an integrative framework defined by P4 (predictive, preventive, personalized, participatory) medicine.^47–49 

Personalized medicine is an attractive concept, however, as with biomarkers it is not new. A physician seeing a patient has always been making a personal judgement of the patient’s status and best ways towards cure, based on visual observation, verbal interaction, physical examination and other personal impressions. What is new, however, is that the molecular analyses of pathophysiology that have recently become possible add to this judgement, enabling rational stratification of the patient and prescription of drugs with distinct mechanism of action to patient subgroups. Taking it one step further, if drug therapies are combined with individualized optimization based on the person’s molecular phenotype, it may become possible to conduct n=1 clinical trials and exercise truly individualized medicine.⁵⁰  Stratified medicine combined with the physician’s personal advice to the patient, e.g. on adapting lifestyle, is already a major step towards that objective.

Cancer treatment and therapies are leading the way in personalized medicine, where its impact is already being observed. With the accumulating molecular knowledge of cancer, it is clear that no two patients – even those with the same diagnosis – experience the same disease. Cancer in itself is a highly heterogeneous disease whereby cells within the same type of tumor may possess different genetic mutations and display different cellular activities and aberrations.^51,52  On the other hand, slight genetic variations in our bodies can alter how our immune systems fight cancer, how rapidly drugs are absorbed and metabolized, and the likelihood and severity of adverse effects. Thus, it is imperative that treatment be designed and prescribed to address the unique biology and molecular phenotype of each patient’s own disease, thus to maximize the chance of successful outcome.

Personalized medicine aims to leverage our molecular understanding of disease to enhance preventive health care strategies while people are still well, thereby advocating preventive and pre‐emptive medicine. The overarching goal of personalized medicine is to optimize medical care and outcomes for each individual, resulting in customization of patient care.⁴³  Towards this goal, the success of personalized medicine will depend on successful deployment of several practices in clinics:

• Pharmacogenomics: whereby genomic information is used to study individual responses to drugs. When a gene variant (called an allele) is associated with a particular drug response in a patient there is the potential for making clinical decisions based on genetics by adjusting the dosage or choosing a more effective drug. Various genetic loci (and their alleles) associated with known drug responses have been identified, which can be tested in individuals whose response to a particular drug(class) is unknown. Modern approaches include multigene analysis or whole‐genome single‐nucleotide polymorphism (SNP) profiles, and these approaches are just coming into clinical use for drug discovery and development. For example, the enzyme CYP2D6, one of a class of drug‐metabolizing enzymes known as cytochrome P450 gene(CYP) found in liver, metabolizes certain antidepressant, antiarrhythmic, and antipsychotic drugs. Extensive molecular characterizations of this gene has led to identification of 70 variant alleles, which are known to exhibit substantial difference in drug metabolism.⁵³  Leveraging this information a diagnostic chip designed by Roche called AmpliChip P450 Array can determine genotypes for alleles of selected CYP genes, including CYP2D6, providing valuable information for appropriate drug administration.⁵⁴ 

• Family Health History and Disease Risk Assessment: Family health history (FHH) is an invaluable and till now rather underutilized information in the practice of personalized medicine. A robust and objective assessment of FHH, which reflect the complex combination of shared genetic, environmental, and lifestyle factors, can inform patients of potential risk factors for certain types of diseases (e.g. Type 2 diabetes, Coronary Artery Disease, Breast/Colon/Lung cancers).^55–57  Genetic testing can provide valuable information by profiling certain disease related genes – to enumerate particular changes or mutations that could render one susceptible to certain disease, for instance, BRCA1/2 for breast and ovarian cancer,⁵⁸  and APOE4 for Alzheimer’s disease.⁵⁹  A timely FHH assessment would help identify persons at higher risk for disease, enabling pre‐emptive and preventive steps, including lifestyle changes, health screenings, testing, and early treatment as appropriate.

• Digital Medicine: While the “omics” revolution has changed the face of biomedical research, the “digital” revolution in the last decade has transformed our daily lives, engendering profound changes to the way knowledge is generated, shared and propagated. The impact of the digital revolution is being clearly felt in healthcare and it is predicted to become one of the key drivers of next‐generation healthcare and personalized medicine. One of the components of the digital ensemble, noninvasive imaging technologies, has revolutionized clinical diagnostics for cardiovascular diseases, various forms of cancers and neurodegenerative disorders. Recently, an imaging approach based on positron emission tomography (PET) with Pittsburg Compound B, called Amyvid, was approved by FDA as a test for diagnosis of Alzheimer’s disease.⁶⁰  Digital medical devices coupled with powerful wireless technologies provide hitherto unavailable options to monitor various aspects of personal health and vital statistics (e.g. heart rate, glucose level, blood pressure) in a continual manner.⁶¹  Electronic health records (EHR) intend to be the backbone of the proposed healthcare system of coming decades and till‐date is a largely underutilized resource. However, with the growing amount of personal medical data being generated as part of modern clinical practices (clinical, omics, imaging, sensors), the role of EHR will be pivotal for the success of personalized and integrative healthcare.⁴⁴ 

The human genome project was a milestone in biomedical research that ushered in the era of molecular profiling (the “Omics”) and opened a new vista for studying biological processes and systems at various levels of cellular organization and hierarchy.⁶²  In this section we briefly review the technology platforms that constitute the molecular profiling toolbox for the exploration of the role of the genome and its expressed products (transcriptome, proteome and metabolome) on health and disease states (Figure 1.5).

The foundations of personalized medical care would be built on rational clinical application of these technologies:

• Genome‐Wide Variation: Clinical characteristic of diseases such as cancer and certain individual response to drug treatment can be attributed to a person's genotype. Propelled by various technological breakthroughs particularly Next‐Generation Sequencing, DNA sequencing has advanced our understanding of multiple diseases by elucidating underlying functional gene variants that are perhaps causal of the disease state, for example in cancer, rare genetic diseases, and microbial infection.^64–66  The use of whole‐genome sequencing has matured to a level whereby it is has been successfully applied in clinical settings for definitive diagnoses and to guide treatment regimen.^67–70  On the other hand, various haplotyping projects have led to identification of SNPs that are the causative variants of disease,^71,72  thereby providing framework for various Genome‐Wide Association Studies (GWAS). While many of these disease‐related SNP associations obtained across over 1400 GWAS studies⁷³  have yet to show clinical relevance, there are a few exceptions. For instance, in a GWAS to see if the drug response to Hepatitis C treatment could be predicted, Ge et al.⁷⁴  identified a SNP at IL28B gene to be associated with a twofold change in treatment response. This finding was later confirmed in a clinical study that genotyped patients receiving treatment for the hepatitis C virus and found the same polymorphism to be a strong predictor of sustained virologic response.⁷⁵ 

• Transcriptomics: For various diseases transcriptomic data generated using microarray platforms have been successfully used for diagnosis, prognosis, and in predicting response to therapy. In oncology, these data have uncovered distinguishable molecular classes for many cancers – including breast, lung, blood‐based lymphomas, leukemia and melanoma^76–80  Similar disease classifications based on microarray data have been performed in other complex diseases, including cardiovascular disease, rheumatic diseases and their inflammatory pathways, and neurologic diseases such as multiple sclerosis.^81–83  Next‐Generation Sequencing technologies such as RNASeq are providing insights into transcripts and their regulation at an unprecedented depth and precision.⁸⁴  For example, transcriptomes relating to cancer have been investigated through RNASeq, providing a new approach for understanding gene fusion and translocation events.^85,86  Owing to advances in genome‐sequencing technologies personalized oncology is becoming a possibility.⁸⁷  In a recent pilot study, high‐throughput sequencing involving whole‐genome sequencing, exome sequencing and RNASeq were employed to guide clinical decisions for patients suffering from metastatic colorectal cancer and malignant melanoma.⁶³ 

• Proteomics: Advances in mass spectrometry (MS)‐based proteomics has enabled high throughput, broader coverage, and accurate quantitation of proteins,⁸⁸  thereby opening newer vistas for diverse systems‐wide proteomic investigations of clinical relevance.^89–91  Consequently, proteomic datasets are getting richer and being applied in understanding of human health and disease. In a recent example, global proteomic profiling was successfully employed to classify diffuse large B‐cell lymphoma subtypes.⁹¹  In another example, system‐wide proteome analysis of breast cancer cell lines recapitulate the disease progression and provided novel prognostic markers for ER(–) tumors.⁹²  Recently, an innovative method of “proteome‐wide analysis of SNPs” (PWAS) was reported. This method enables rapid screening of SNPs for differential transcription factor (TF) binding and was successfully applied to elucidate differential TF binding to type 1 diabetes‐ (T1D‐) associated SNPs at the IL2RA or CD25 locus.⁹³  While still an emerging discipline, continued advances in mass spectrometry (MS)‐based proteomics in combination with innovative experimental strategies, and advances in computational methods will surely play a profound role in personalized medicine.⁹⁴ 

• Metabolomics: Study of changes in nonprotein small‐molecules metabolites could reflect biological and (patho)physiological state associated with disease. Metabolomics has been applied to an array of diseases, including diabesity, cardiovascular disease, cancer, and mental disorders.^95–98  In drug discovery metabolomic profiling has led to identification of enzymatic drug targets and as a tool for assessing drug toxicity.⁹⁹ 

Biomarkers are at the core of personalized medicine as they are the instruments by which clinicians decide which patients to give what drug at what dose and time. The impact of biomarkers on personalized medicine varies between therapeutic fields but encouraging results have started to emerge of patient selection for tailored therapies, for instance in cancer. Cancer is largely a genomic disease that manifests due to a spectrum of genomic abnormalities acquired by normal human cells, ranging from point mutation, structural variations (chromosomal rearrangement) and gene fusion and genomics biomarkers have shown attractive applications for patient selection.¹⁰⁰ 

At present, 103 drugs are available whose label contains pharmacogenomic biomarker information, which drives the prescreening of patients to maximize drug efficacy or safety.¹⁸  A total of 113 drug–biomarker combinations are characterized in multiple disease areas of which the main applications are in oncology (29%) and psychiatry (24%) (see Figure 1.6).

Roughly half of these drug–biomarker combinations relate to genetic Cyp450 family variants, mostly Cyp2C19 and Cyp2D6, that affect the metabolism of the drugs. On the one hand, this could result in too high (or too low) compound levels in poor (or ultrarapid) metabolizing patients, requiring dosing adjustment to have sufficient active compound present whilst not inducing safety liabilities. For psychiatry, multiple examples of this exist, reviewed by Zhang et al.¹⁰¹  On the other hand, Cyp450 metabolism may be required to convert the inactive parent drug to its active metabolite form, as is the case for the conversion of the antibreast cancer drug Tamoxifen to its active metabolite 4OH‐Tamoxifen by Cyp2D6.¹⁰² 

The other part of the drug–biomarker combinations relates to biomarkers whose presence determines the efficacy of the drug in that patient. Particularly in oncology, there are striking examples. One of early examples of individualized cancer treatment based on genetic status of an individual was demonstrated in the case of Herceptin (Trastuzumab), a monoclonal antibody specifically used in treatment of early‐stage and metastatic HER2‐positive breast‐cancer patients.¹⁰³  Based on data correlating efficacy to expression, FDA requires profiling of the tumor cells from breast‐cancer biopsies to check for increased HER2 expression before Herceptin administration.

Alternatively, certain gene mutations are “driver mutations” that result in an overactive signalling pathway and strong tumor growth. An example is the BRAF^V600D/E point mutation that is found in ∼60% of patients with malignant melanoma.¹⁰⁵  This lethal form of cancer reacts poorly to conventional chemotherapy and radiation treatment, resulting in average survival rate of less than a year after diagnosis. Recently, an orally active BRAF‐mutation‐directed drug vemurafenib was developed with outstanding results, more than 80% of patients responded to this drug with tumor shrinkage in just 2 weeks. This effect was only seen in patients carrying the specific mutation and those without the mutation responded with worse outcomes upon treatment.^104,106  Other examples of drug–biomarker combinations include gefitinib–EGFR mutations¹⁰⁷  and crizotinib–ELM‐ALK4 gene fusion¹⁰⁸  for treatment of nonsmall cell lung cancer.

In view of the potential roles of biomarkers as discussed above, it is interesting to reflect on the trends, challenges and opportunities regarding biomarkers in translational and personalized medicine in the pharmaceutical industry.

Within pharmaceutical research and development, biomarkers have matured from being a hype to becoming an intricate part of the decision‐making process. The low success rate and high costs associated with development of a novel drug has pushed pharmaceutical companies to try their utmost to be more efficient and cost effective. One of the strategies is to focus on biomarkers. Rather than trial‐and‐error, most if not all drug‐development projects have key biomarker read‐outs incorporated in their strategy to enable early derisking of the chosen approach and selection of the best compound moving forward. In parallel and in line with the desire for personalized medicine from society, there is a clear trend with oncology leading the way that new drugs have a companion diagnostic biomarker where possible.

A clear trend in the pharmaceutical field has been the externalization of biomarker research. For both standardized and specialized activities it was found to be more cost‐, resource‐ and time‐effective to stop doing these activities inhouse and to outsource such tasks to dedicated service providers that perform high‐quality contract research as their core business. As a direct consequence, many activities within companies and universities are currently being outsourced, creating interdependent functional networks. Such functional networks have a global nature and topology in which distance is not a key factor, but rather the quality, speed, costs and professional skills of the external partner. Many scientists nowadays have developed global connections and routinely send samples across continents for analysis by contract research organizations or academic collaborators.

The marketplace for pharmaceutical drugs has also globalized and with it, the research into their effectiveness and safety has taken a globalized view. Whereas earlier most, if not all, of the commercial and research focus of pharmaceutical and diagnostic industries was on Western markets in Europe and USA, recently this has shifted to the emerging markets, spearheaded by Brazil, Russia, India and China. However, distinct differences have been found in response to drugs originally developed for the Western markets, but also in the disease pathophysiology. For instance, the prevalence of certain cancers such as gastric and lung tumors is highly increased in Asia, caused by genetic and environmental risk factors that are different from those in western patients such as infection with hepatitis B or hepatitis C virus.¹⁰⁹  Also, South Asians from the Indian subcontinent have an increased risk of developing coronary heart disease as compared to Europeans, which is postulated to originate in different risk factors including insulin resistance, type 2 diabetes mellitus, lower physical activity, diet, in combination with a different coronary anatomy. Specific correlating biomarkers including blood levels of lipoprotein(a) and homocysteine, and candidate genetic markers were identified.¹¹⁰  The global variation in disease prevalence and treatment success is due to variable exposure to risk factors, only some of which are known, making personalized medicine across the globe a real challenge but at the same time an opportunity. Without doubt, the different genetic backgrounds, dietary preferences and lifestyle of populations across the globe contribute to this, but clearly more research is needed to investigate these aspects.

Supported by their increased economic power, the emerging countries currently invest heavily in biomarker research, aiming to improve existing therapeutic treatments and to generate more personalized drugs for their markets. To facilitate such research, large science parks with impressive infrastructures have been set up, such as the Biopolis in Singapore and the Zhongguancun China Science Park in Beijing to name just two. These infrastructures bring together public and private partners to facilitate interactions towards applied research and development.

Methodological developments have strongly increased the potential for identification of new biomarkers. First, new data‐rich biomarker discovery technologies have emerged that dramatically increased the amount of molecular information of preclinical and human systems. Particularly, next‐generation sequencing technologies have revealed new possibilities to identify variants in the DNA and RNA species. Besides the genetic code itself, epigenetic regulation through methylation and acetylation, synthesis and composition of RNA species including messenger RNA and micro‐RNA on nucleotide level can now be studied.^84,111  The costs of next‐generation sequencing has dramatically decreased, enabling application of genome and whole exome sequencing in clinical diagnostic research to identify genetic causes of unknown diseases.¹¹²  Also, mass‐spectrometry methodologies have strongly developed and enabled the analysis of changes in peptide, protein and metabolite isoforms that reflect perturbations of biological systems. Complementing the sequencing analysis, mass spectrometry has yielded a wealth of observations on variants of the proteome and metabolome in healthy and diseased states.^113,114 

Biomarker discovery laboratories increasingly use multiple complementary molecular profiling approaches in parallel to identify a biomarker that fulfils the objective of a study.¹²  This strategy not only increases the chance of identifying candidate biomarkers but it also generates independent data that provides additional mechanistic support of the best biomarker. As such, various molecular profiling combinations have been reported including of proteomics–metabolomics,¹¹⁵  proteomics–lipidomics,¹¹⁶  genetics–proteomics,¹¹⁷  genetics–metabolomics,¹¹⁸  and transcriptomics–proteomics.^90,119,120  Recently, considerable progress has been made in the development of novel chemometrics and bioinformatics methods to integrate biomarker‐profiling data from the different platforms with phenotypic observations.¹²¹ 

This progress has dramatically changed the landscape of biomarker discovery and strongly increased the potential of identifying more specific and applicable biomarkers for human medicine.

The multitechnology biomarker‐discovery approaches described above are likely to lead to biomarker panels rather than single biomarkers, as a panel of functional biomarkers read‐outs is likely to have improved predictive value over single entities. However, it is not always practical to develop a biomarker panel test and the optimal condition for one test may be suboptimal for another in the same panel. Consequently, there is a need for simple tests that have an easy read‐out, require minimal sample amounts and can be produced at low cost. Such assays can then be applied in single‐plex format, but the data can be combined in multiplex format to have the desired multiparameter comprehensive view.

A related trend is to combine quantitative molecular biomarker data and clinical read‐outs to provide a more detailed characterization of a disease state. This enables clinicians not only to diagnose diseases but also to specify the subtype or causal origin of the pathophysiology, potentially leading to a more tailored treatment. A nice example of this is the Disease Activity Score (DAS28) as applied in characterization of rheumatoid arthritis patients (see www.das28.nl/das28/en/). The DAS28 assessment is composed of a physical examination of the patient joints (the number of swollen and tender joints among 28 vulnerable joints), blood biomarker analysis (erythrocyte sedimentation rate) and a patient self‐assessment. The DAS28 score will yield a number between 0 and 10, indicating the status of the rheumatoid arthritis disease at that moment.

At present, biomarker scientists have become particularly efficient in collecting enormous amount of data at the whole‐genome level. The translation of this wealth of data to knowledge and clinical insights, however, is still a big challenge. Improved statistical, bioinformatic and chemometric analysis methods are needed to smartly mine the high‐content data and filter out the signals that can be used for identification of biomarkers and/or put them in a proper functional context.

Importantly, and a related challenge, many novel variants of genomic, transcriptomic, proteomic and metabolomic nature are identified, whereas knowledge on their biological function is scarce, resulting in a large number of potential new biomarkers that require further studies and validation. In this regard a new breed of clinical investigators and translational life scientists would need to be trained who are eager to embrace the breakthroughs in molecular‐profiling sciences and amalgamate that with advances in medicine – to drive decisions for tailored treatment and therapies thereby advancing personalized medicine.^122,123 

A major challenge in the biomarker field today resides in clinical biomarker validation. Even with the advances in molecular‐profiling technologies we are still left with very few clinically accepted biomarkers. This discrepancy can be leveled by employing effective translational medicine approaches leading to expedited “bench to bedside” results. However, human diseases are mostly complex in nature and it requires several biomarkers to mechanistically describe the imbalance in the metabolic equilibrium responsible for the disease and the effect of treatment thereon. The number of published biomarkers with clear biological relevance needs to increase and, correspondingly, the clinical validation of such biomarkers needs to improve strongly. Without extensive validation, the outcome of a biomarker test cannot be used in driving important clinical decisions such as in selecting patients for personalized therapy and/or treatment. Clinical biomarker validation should optimally be performed by applying standardized procedures and protocols to test independent and well‐characterized clinical samples across multiple independent laboratories. However, in reality, this is rarely carried out. The commitment of biomedical researchers to engage in long‐lasting and expensive biomarker‐validation projects is limited, which is at least partly due to the pressure to publish innovative findings in high impact journals and secure funding. Although clinical validation of biomarkers, including the development of quantitative biomarker assays with high accuracy, specificity and reproducibility is essential to enable translational and personalized medicine, it has limited new scientific value and as such is less prone to obtain funding and high‐impact publication. Worryingly, the initial effect size of a biomarker in selected clinical populations can in the majority of cases not be confirmed in meta‐analyses of subsequent studies.¹²⁴  Several reasons can be identified for this observation including differences in study designs, study sizes and subject inclusion, small variations in the biomarker analytical protocol used, differences in the isolation and preparation of the new clinical samples versus the ones on which the original finding was based, and potentially the biased selection of most differential data to publish the initial biomarker finding. This worrisome observation hampers the implementation of biomarkers by others and blocks the progression of the field at large.

To be able to mature to a molecular diagnostic test, candidate biomarkers need to be developed through a rigorous process. This starts with demonstration of the added value in clinical practice, followed by development of a prototype diagnostic test, its thorough analytical and clinical validation by several independent laboratories, demonstration of its positive health economic value, alignment with regulatory and insurance agencies to obtain registration and reimbursement of the test, and finally implementation by clinicians.^19,125  This lengthy process is aimed to optimize the specificity and selectivity of the diagnostic test, whilst minimizing its cost and demonstrating its usefulness in clinical practice. Whereas great progress has been made in the front end of this diagnostic development pipeline, little progress has been made in the latter part. In particular, major gaps remain in the diagnostic test development and multicenter validation of candidate biomarkers, as this type of work is generally not easily funded by grants. A few positive exceptions exist such as the recent multicenter testing of the key Alzheimer’s disease biomarkers Aβ42, T‐tau, P‐Tau.¹²⁶ 

Despite constructive efforts by regional biobanks, there is an intrinsic lack globally of (pre)clinical samples that are highly characterized, are properly stored, have associated patient phenotype metadata and are being made available for multicenter biomarker validation. Consequently, lack of such biospecimen can lead the researchers to focus on objectives that are considered more achievable in the short‐term rather than what is desired in the specific field in the long term. Hence, biomarker studies may be statistically underpowered, only rarely include reporting of robust clinical biomarker validation with sufficient clinical sample size, and hence may not be reproducible by other laboratories. It is imperative for high‐quality biomarker data that high‐quality biosamples are tested, but it will be a challenge to organize the global community so as to enable this.

Successful biomarker discovery, validation and development requires a smooth and cooperative interaction between key players in this field including patients, academia, method and instrument developers, the pharma industry, the diagnostic industry, clinicians, health‐insurance providers, regulatory approval agencies, government policy makers, and many others. Although seemingly logical, such across‐the‐board collaborations only rarely exist, for various reasons including different interests and objectives, unawareness of the potential synergy of working together, lack of funding and others. In the current economic climate, securing funding is a major challenge. The consequence is that too few biomarkers progress from discovery to clinical implementation and are stalled at intermediate stages. The challenge here is to bring those key players together in a mutually beneficial working environment with sufficient funding of the joined research to enable the pipeline to be operational.

An important component of application of biomarkers in real life, is to identify and understand the variation in biomarker levels across various population groups locally and in the world. Some of these can be related to variables in genetic background, lifestyle or nutritional intake, but many cannot. A greater understanding of this natural variation is imperative prior to the use of such biomarkers to mechanistically characterize disease and predict treatment effects.

Despite their potential in clinical settings to drive diagnosis and treatment strategies, unanswered questions still remain about the reimbursement of costs associated with expensive biomarker tests. Health‐insurance companies generally do not want to pay for diagnostic tests that are not part of a hitherto established medical diagnostic toolkit. This has particularly become an issue with genomic biomarker screening that can be key to enable patient selection for rational treatment, as also recently exemplified in the case of Kalydeco to screen for a rare mutation of CFTR gene to maximize the success of cystic fibrosis treatment.¹²⁷  The cost of whole‐genome analysis of an individual has dramatically decreased to about USD 1000, but it is yet unclear who should pay for these multibiomarker tests that will lead to one selected preferred treatment.

The deluge of molecular biomarker and profiling data raises a key ethical question related to data privacy, access and use.^128,129  For any individual, having access to his/her own medical data through electronic health records could certainly boost their participation in disease management. However, the knowledge of having an identified genetic susceptibility to a certain disease may cause serious disruption of the person’s peace of mind, aggravating the occurrence of disease symptoms. Even worse, if such data becomes available it may have serious implications for the individual’s relation with employers and health‐insurance companies. Who will have access to genetic information and how can we ensure that this information would not be wrongly used and interpreted? The challenge here is to define this properly so as to safeguard the individual’s identity and interests.

Embracing an optimistic view, we feel that every challenge is an opportunity for progress and the challenges outlined above translate into specific objectives for improvement, both scientifically and process‐wise. We would like to believe that the scientific improvements, such as those dedicated to improved data‐analysis methods and improved understanding of the biological background of novel biomarkers, will be addressed by the innovation‐driven research of academic and pharmaceutical researchers. The process‐wise improvement in biomarker research and development, however, requires a different mind‐set and working model. The desired pipeline of biomarker discovery–validation–development–implementation requires that multiple parties work closely together with the objective to apply known biomarkers rather than discover new ones.⁴⁸  Such an application‐driven working model is ideally suited for public–private partnerships that bring together academic innovation, specialized technical skills, clinical expertise and industrial process focus.

The downsizing of internal biomarker activities in larger companies has already resulted in a considerable stimulation of new economic business opportunities. Many specialized spin‐off companies have been formed by entrepreneurial scientists who excel in a specific part of the biomarker research and development process. Together with contract‐research organizations, technology vendors and central analysis laboratories, they form a functional network with larger companies and universities to drive pharmaceutical and diagnostic biomarker development. Importantly, this functional network now also includes the emerging markets and their populations, driving biomarker research across the globe.

Although across‐the‐board funding opportunities have decreased, funding agencies generally prefer to support multicenter consortia rather than single laboratories. Good examples of public–private partnerships that we have recently been involved in include Netherlands programs such as Top Institute Pharma and Center of Translational Molecular Medicine, USA-based Critical Path Initiatives and European framework programs. These partnerships have brought together basic and applied biomarker researchers in academia and industry and have stimulated knowledge transfer and a focus towards applied biomarker research. Public–private consortia have been shown to lead to increased awareness of the value of crossdiscipline biomarker research, improved data‐analysis workflows, and increased identification and development of biomarkers for specific diseases and mechanisms.

The opportunity now lies in the formation of a large public–private partnership that is fully focused on the validation and development of biomarkers, aiming at clinical implementation. We believe the mind‐set of key stakeholders is currently set to form and support such functional network. This should bring together pharmaceutical industry, academia, clinical experts, biobanks, analytical laboratories and diagnostic industries in a working model that creates a win‐win situation for all. Pharmaceutical industries have a key interest in applying validated disease‐related biomarkers and can contribute clinical samples from their intervention trials. Academic researchers can provide indepth knowledge on the mechanistic relationship of a biomarker with disease. Clinical experts are imperative to indicate the biggest clinical unmet need, thus focusing the network on the clinical application, and by making sure the right biosamples are included in the clinical validation projects. Biobanks have a platform to contribute their unique clinical samples to, thus obtaining compensation for their investments while maximizing the use of their samples to generate high‐impact data. Analytical laboratories can bring in their specialized skills in biomarker‐assay development and analytical validation and, where needed, contribute to the registration of the biomarker as a novel diagnostic product. Diagnostic companies can further develop the clinically and analytically validated biomarker test and convert it into a commercial product. The scale of such public–private networks should be large enough to enable independent cross-laboratory validation of a biomarker test, demonstrating its added value in a particular application or, perhaps even more importantly, its lack thereof due to insufficient robustness. Reporting of those positive and negative validation results are crucial to advance the field.

In parallel with well‐defined biomarker development projects, academic researchers and specialized methodology scientists can contribute to this network by further improving biomarker technologies where needed. Through benchmarking of novel technologies versus established ones in biomarker development projects, the network can demonstrate their added value, thus facilitating acceptance of methodology improvements.

Looking ahead, focused biomarker development networks as outlined above are likely to boost the quality of published biomarkers and their application in translational and personalized medicine. Ultimately, this higher level of biomarker knowledge and tools will result in transformation of disease management into health management. At present, it may seem a utopia but with the impressive progress we have seen in recent years it may occur sooner than expected.