Welcome to the era of Genomic and Personalized Medicine. This article will tell you what you need to know and how to be best armed to take full advantage of emerging genomic information and incorporate it into the daily practice of medicine.

Medical Practice Circa 2015

Consider this case example of how genomics might be integrated into medical practice within the next 10 years:

Jackie, a 50-year-old woman, had her genome scanned 4 years ago, in 2011, by her physician and was found to have 5 genetic variants that increase her risk for heart disease threefold; she also had 3 genetic variants that confer protection against cancer. Recommendations for her diet and medications have been made on the basis of her genetic profile for metabolizing nutrients and specific medication classes. During her last examination, she underwent a screening blood-based profile for coronary artery disease that not only showed that her atherosclerotic disease burden was high but also that her overall risk for developing an acute coronary syndrome in the next 5 years was high. When chest pain developed recently, a blood-based molecular biosignature confirmed ischemia caused by intermittent thrombosis without necrosis. Targeted anti-inflammatory and individualized dosing regimens for anticoagulant medications were prescribed and symptoms did not recur.

Background

Medicine today already contemplates the use of molecular markers to predict health risks. Glucose and cholesterol measurements, for example, are standard of care. However, whereas these are good screening tests for occult disease, they are not necessarily 'predictive' of future events. Genomic information provides the opportunity to develop and refine the prediction for risk and increase the precision of that prediction for the individual. Sir William Osler (1849-1919) recognized that "variability is the law of life, and as no two faces are the same, so no two bodies are alike, and no two individuals react alike and behave alike under the abnormal conditions which we know as disease."^[1] We now have a new set of tools that can be used to understand biological and disease variability. However, health and disease management today are further complicated by a pharmaceutical industry that has developed medications using a one-size-fits-all paradigm rather than medications tailored to human variability. Thus, in medical practice today, many drugs work in fewer than 50% of the patients to whom they are prescribed. Furthermore, more than 100,000 people die annually from drug-related adverse events -- a top 10 cause of death in 1994.^[2] In Britain, it was recently estimated that adverse drug reactions are responsible for 6.5% of that country's hospital admissions and for nearly $1 billion in annual costs to the National Health Service.^[3] A more personalized or customized approach to medical care might be part of the solution to these healthcare woes.

What Is Personalized Medicine?

Personalized medicine is a rapidly advancing field of healthcare that is informed by each person's unique clinical, genetic (DNA-based), genomic (whole genome or its products), and environmental information.^[4] The goals of personalized medicine are to take advantage of a molecular understanding of disease to optimize preventive healthcare strategies and drug therapies while people are still well or at the earliest stages of disease. Because these factors are different for every person, the nature of disease, its onset, its course, and how it might respond to drug or other interventions are as individual as the people who have them. For personalized medicine to be used by healthcare providers and their patients, these findings must be translated into precision diagnostic tests and targeted therapies. Because the overarching goal is to optimize medical care and outcomes for each individual, treatments, medication types and dosages, and/or prevention strategies may differ from person to person -- resulting in unprecedented customization of patient care.

Specific advantages that personalized medicine may offer patients and clinicians include the following:

• The ability to make more informed medical decisions;
  
  
• Higher probability of desired outcomes thanks to better-targeted therapies;
  
  
• Reduced probability of adverse side effects;
  
  
• Focus on prevention and prediction of disease rather than reaction to it;
  
  
• Earlier disease intervention; and
  
  
• Reduced healthcare costs.

What is Genomic Medicine?

Genomic medicine is an essential component of the broader personalized medicine concept. Simply defined, it is the use of genomic information to guide medical decision making.^[5] The prospect of examining a person's entire genome (or at least a large fraction of it) to make individualized risk predictions and treatment decisions is tantalizingly within reach. Since the completion of the sequence of the human genome, genomic information has been used, albeit on a small scale, to tailor care to individual patients. The opportunity, however, is enormous: for the first time, we are in a position to characterize health and disease states by their molecular fingerprints, develop meaningful stratifiers for patient populations, elucidate mechanistic pathways based on genome-wide data, and develop new preventive, diagnostic, and therapeutic strategies that will shift the focus of care from intervention to prevention.

What Is the Human Genome and How Does it Vary From Person to Person?

The human genome is composed of approximately 25,000 genes defined among 3 billion bases or nucleotides of DNA residing on 46 chromosomes. Variation in this sequence from one individual to another can occur even at the level of a single base (termed a single nucleotide polymorphism [SNP]). Other types of variation include deletions, insertions, repetitive elements, and large chromosomal rearrangements. All of these may have medical implications, with the latter (deletions, insertions, repetitive DNA elements, and large chromosomal rearrangements) being responsible for a large number of congenital malformations and inherited single-gene diseases. The former -- the SNPs -- largely underlie complex human diseases and their contribution to these diseases has been among the most challenging roles to define. A polymorphism is a genetic variant that has 2 alleles in a population at a frequency of > 1%. SNPs are the most common form of genetic variation and there are an estimated 10,000,000 SNPs in the human genome. High-throughput technologies such as direct sequencing, various genotyping platforms, and microarrays enable rapid and accurate analyses from very small quantities of blood or tissue samples. SNPs are classified as non-synonymous if they cause an altered protein amino acid sequence and synonymous or neutral if protein amino acid sequence is unchanged. They may also fall in an intergenic region (eg, intron, promoter, untranslated region). Any of these may lead to disease caused by changes in protein structure or function as a result of amino acid changes or alternative splicing or by changes in protein quantity as a result of altered transcription factor-binding sites or other regulatory disruptions.

Genetic variation is complex at a population level, with many SNPs existing in different populations (of different ethnic origins) and at different frequencies. To simplify which SNPs should be measured in various populations, the International Hap Map Project was started in 2002 as a multicountry effort to identify, catalogue, and make publicly available the genetic diversity in humans.^[6] The HapMap has identified which SNPs travel with others in blocks of the genome (a phenomenon termed linkage disequilibrium) such that the measurement of one SNP is informative for a large number of neighboring SNPs that therefore do not have to be directly measured. A complete version of the HapMap can be found at http://www.hapmap.org. The HapMap project has been important in facilitating genetic discovery -- investigators can now focus on those SNPs that are informative about a series of SNPs in a single block of the genome. This reduces the number of SNPs needed to test for genetic studies, increasing the speed of the studies and decreasing their cost.

SNPs are increasingly being used in clinical practice for the diagnosis of Mendelian disorders such as cystic fibrosis and long QT syndrome. However, there are few, if any, SNPs used in clinical decision making for complex diseases such as asthma or autism. The multigenic nature of these common diseases, the modest effect of the genetic variant(s), and the confounding effects of environmental influences have made them exceedingly difficult to validate in multiple populations. A predictive genetic test for a complex disease would likely consist of a panel of multiple SNPs combined with other genomic and clinical information.^[7]

What Are the Types of Genomic Information?

Having access to the entire human sequence is a necessary but insufficient prerequisite for genomic medicine. What is equally important is having the technology at hand to reliably visualize individual genomes (as well as their derivatives, the transcriptome, proteome, and metabolome) for health and disease status (Table).

Table. The Genomic Medicine Toolbox

SNP = single nucleotide polymorphisms

Gene Expression Variation (Transcriptomics). Gene expression profiles can also serve as important markers for disease and clinical outcomes. Gene expression profiles generated from microarrays of more than 22,500 transcripts characterize overall transcriptional activity in a particular cell type as it is affected by gene sequence variation, regulatory controls, and environmental influences. Gene expression patterns as a whole are commonly used in classification and predictive models to differentiate tumors that are similar histologically. Microarrays -- small, solid supports onto which the sequences from thousands of different genes are immobilized, or attached, at fixed locations -- are the major tool used to measure genome-wide changes in gene expression; however, older platforms such as reverse transcriptase polymerase chain reaction are routinely used for smaller gene sets. This approach has been used experimentally for classifying acute myelogenous leukemia subtypes, for distinguishing Burkitt's from diffuse large B-cell lymphoma, for predicting outcomes in breast and lung cancers, and for assessing atherosclerotic burden in human aortas.^[8-14]

Although patterns of gene expression composing a molecular profile or signature that correlates with a phenotype can be observed, it is often unclear what the relationship is between the pattern and the cellular pathways or networks relevant to the disease process in question. For example, an expression profile of 16 cancer-related genes has been associated with the risk for breast cancer recurrence, but it is not yet clear how each of the genes function, whether separately or together, in the pathophysiology of breast cancer.^[14] A noninvasive molecular expression profile has also been developed and validated to categorize heart transplant patients into low- and high-risk categories for cardiac rejection before clinical symptoms appear^[15] Again, the pattern is predictive, but the role of the genes in allograft rejection is not yet known. Recently, genomics-based assays have been released commercially to aid in medical management of breast cancer treatment^[16] and cardiac transplant.^[17]

Protein Expression (Proteomics) Assays. Proteomics involves the large-scale study of the structure, quantity, and function of proteins and provides a comprehensive dataset about cellular networks, interactions, and products. Proteomics naturally followed the field of genomics to characterize the protein make-up of cells in different tissues, organisms, and disease states.^[18] The proteome is estimated to contain approximately 100,000 protein entities and their post-transcriptionally modified derivatives. One of the major challenges to this young field is distinguishing protein signatures linked to a specific drug or disease from background caused by confounding responses to environment, lifestyle, or comorbidities. While still in its early days for diagnosis and prognosis, multiplex proteomic assays used to detect a profile of multiple proteins are being developed for disease management. A combined measure of C-reactive protein, B-type natriuretic peptide, and cardiac troponin I can be used for risk stratification of patients with acute coronary syndrome^[19] and is the harbinger of multiplex proteomics diagnostic and prognostic assays to come.

Metabolic Profiling (Metabolomics). Another rapidly emerging technology is metabolomics. Metabolomics is the analysis of low molecular weight molecules involved in primary or intermediary metabolism and metabolites found in cells, tissues, and body fluids. The metabolome represents the culmination of all molecular events in a specific tissue -- gene variation and expression (genomics), protein expression, modification, and activity (proteomics), and environmental exposures.^[21] A conservative estimate of the number of metabolites is 5000 compared with the more than 100,000 estimated protein species and more than 20,000 genes in humans. Because the metabolome represents an end stage of cell events, it may correlate with phenotype more directly.^[20] However, this strength may also be a weakness because the metabolome is likely to be extremely dynamic and sensitive to a multitude of signals, disease-related and otherwise.

Single metabolite data are quite familiar to clinicians, with serum cholesterol and serum glucose as prime examples. Advances in high-throughput mass spectroscopy or nuclear magnetic resonance technologies allow detection, analysis, and identification of a range of metabolites (entire classes of small molecules, such as fatty acids, amino acids, nucleosides, and vitamins) that traditionally have been measured one at a time. The resulting metabolomic patterns are being developed as risk and monitoring markers.

Because all of these technologies are genome scale and high-throughput, patterns of biomarkers instead of measurements of single entities are becoming more commonplace. The benefit of detecting multiple measures of change is the ability to view the downstream biological events in aggregate, yielding a more complete picture of the disease potential, progression, or prognosis earlier in the process.

Genomic Tools for Prediction and Personalized Care

How Do They Fit Together?

Emerging know-how and technologies from the human genome are enabling predictive and proactive approaches across the spectrum from health to disease (Figure). Disease susceptibility and risk can now be quantified and anticipated during health and even at birth using stable genomics or DNA-based approaches that do not change over a person's lifetime. Individual SNPs or multi-SNP panels of genes are emerging that might be used as part of health risk assessment (see below). DNA variation also provides information about the possibility of being relatively protected from disease development as well as information about one's sensitivity or resistance to certain medications and ability to metabolize nutrients in our diets. All of this can and probably should be done early in life such that a course or strategy to maintain health can be charted well in advance of the development of potentially detrimental lifestyle habits and exposures. The other -omics that are dynamic and interact with and respond to environmental stimuli, lifestyles, diets, and pathogens are rapidly improving capabilities to predict and intervene at an individual level. Transcriptional profiles, protein expression, and levels of metabolites combined with dynamic imaging modalities will provide more precise ways to screen individuals who are at high risk for disease to find the earliest molecular manifestations while the disease is subclinical. This same information may provide a definitive diagnosis and a molecular classification that foretells prognosis. For example, today a HER/2neu-positive breast cancer ascribes that patient to a more aggressive form of the disease and directs care to a much different course than a HER/2neu-negative cancer. Similarly, the selection of drugs can be guided both by the patient's underlying genetic makeup as well as the molecular architecture of the disease in the individual. Given that the evolution of a disease from baseline risk often occurs over many years, healthcare providers must focus on strategic health planning and disease prevention during the most cost-effective times of the disease life cycle to shift the current paradigm of care from disease treatment to personalized care.

Figure. Use of genomic markers to predict, prognose, diagnose, treat, and monitor health and disease. The red line indicates the course of disease from health through death. The black dashed line is the 'symptom horizon' below which the disease remains subclinical. The red dashed line is the desired outcome of an intervention to reverse disease.

Genomics to Identify Disease Susceptibility

Although genetic testing for Mendelian disorders such as cystic fibrosis, Huntington's disease, familial breast cancer, and phenylketonuria, among others, was widely available prior to the genomic era, the genetic basis for complex disease remains unclear. From 1980 through 2002, fewer than 10 genes were associated with complex diseases in humans in contrast to more than 1300 genes that were associated with Mendelian disorders from 1980 through 2001.^[22] Testing for Mendelian disorders has been essential to understanding both the genetic basis of disease and the clinical impact of identifying risk prior to the onset of disease. However, Mendelian disorders are rare and for genetic testing to be widely applicable, the genetic basis for complex disease needs to be understood.

Until recently, techniques to identify susceptibility genes have been limited to linkage analysis and association studies. If genetic risk factors are present, studying families with affected sibling pairs provides a stronger genetic effect in these families because these siblings will likely share genetic regions that underlie the disease phenotype. Linkage studies take advantage of this to identify regions of the genome that are more strongly associated with disease than by chance by looking for microsatellites that are more prevalent in affected sibling pairs compared with controls. Microsatellites are short segments of DNA that contain repetitive sequences of nucleotides. Linkage studies have had limited success. For example, of 9 linkage studies trying to identify susceptibility genes for coronary artery disease, only 4 genes were identified and 1 locus (2p11) was replicated.^[23] Reasons for the limited success of linkage studies include the polygenic nature of complex diseases, with each gene contributing only a small risk to the phenotype; the relatively low level of heritability of complex diseases compared with Mendelian disorders; and underpowered study designs.^[24]

In contrast to linkage studies that are unbiased, association studies look for an increased frequency of a particular genotype at a candidate gene locus in cases compared with controls. In these studies, the candidate genes must be known a priori and are therefore limited by understanding of the genes that contribute to a particular disease. Association studies have been abundant in the literature. For coronary artery disease alone, association of 96 polymorphisms in 75 genes has been reported.^[25] However, aside from the initial limitation of a priori knowledge, genetic association studies have been limited by their lack of reproducibility. Even though the contribution of these types of association studies remains uncertain, it has been suggested that common genetic variants may contribute to common diseases, supporting the role for continued association studies.^[26] Although appealing, the candidate gene approach has been fraught with design issues, including strict adherence to the definition of the phenotypes, adequate sample size, issues of multiple testing, and lack of replication.^[27]

Today we are in the era of the 'genomic gold rush,' with genome-wide association (GWA) studies using high-density genotyping technologies that allow for assays of 500,000 to 1,000,000 SNPs per individual at relatively low cost.^[28] The identification of millions of SNPs, development of high-throughput sequencing methodologies, completion of the HapMap, and creation of large genotyped cohorts have made GWA studies possible. The results in the past year have been no less than astounding, with genetic loci being identified for many complex diseases, including breast cancer, coronary artery disease, myocardial infarction, obesity, diabetes, and prostate cancer.^[28]

These are indeed encouraging data, many of which have been replicated. GWA studies look for an increased frequency of SNPs distributed throughout the genome in cases compared with controls. The use of genome-wide SNPs means that little to no a priori knowledge of genes contributing to an outcome is needed. The completion of the HapMap means that fewer SNPs need to be genotyped to characterize genomic variation, facilitating the use of technologies such as high-density SNP microarrays. Finally, large genotyped cohorts mean that results in one population can be validated in other populations. Thus far, GWA studies have produced robust and reproducible findings. For example, 3 GWA studies for coronary artery disease have been published using 3 different populations, and all have identified a locus at 9p21.^[29,30] This region is not known to contain coding sequence and therefore may be important in helping further our understanding of the molecular basis for coronary artery disease. Type 2 diabetes has been extensively studied, and independent genome-wide scans have identified several loci for diabetes susceptibility: CDKN2A/CDKN2B, CDKAL1, and IGSF2BP2, as well as confirming TCF7L2, PPARG, and KCNJ11, which had been previously identified by other methods.^[31-35] A susceptibility gene for obesity, FTO, was identified by the same groups studying type 2 diabetes.^[36] Genes for Crohn's disease, rheumatoid arthritis, adult macular degeneration, and prostate cancer all have been identified in the past year using genome-wide approaches.^[37-40]

As a result of the above types of studies, large numbers of additional susceptibility markers are sure to emerge in the coming years. However, before these markers can be applied to practice, their clinical utility must be demonstrated; that is, the impact that using the markers might have on health outcomes. First, markers must be shown to provide additional estimates of disease risk over current clinical models. The use of any one SNP for screening complex diseases such as cardiovascular disease has only a minor probability of providing much of a predictive or correlative significance. However, if combined, multiple SNPs, each with only a small predictive value, might provide enough power to be clinically significant.

In addition, we need to know what to do with the results of genetic tests and define actionable options that our patients can take with these results in hand. Screening for markers of susceptibility offers the unique opportunity for the prevention of disease prior to the onset of clinical manifestations or mitigation of the clinical course of disease. One type of intervention includes lifestyle modification. For example, adherence to a low phenylalanine diet in neonates identified to have phenylketonuria can lead to normal brain development. One could hope for a scenario where knowledge of susceptibility to coronary disease or chronic obstructive pulmonary disease may facilitate smoking cessation. Preliminary data suggest that a personalized approach to smoking cessation improves quitting.^[41] Another type of intervention includes aggressive screening programs to identify preclinical disease. This has been the approach used in patients with pathologic mutations of BRCA1 and BRCA2, although there are few data to suggest that intense screening reduces mortality in these patients. For rare disorders, consensus opinion may be sufficient to justify such an approach, but for more prevalent disorders and susceptibility markers, outcomes data will be needed to provide justification for an intensive screening approach. This has important ramifications for the types of susceptibility markers that are used clinically: markers must be prevalent enough to select enough individuals to adequately power outcomes-based clinical trials. A third type of intervention is either curative or prophylactic therapy, such as prophylactic thyroidectomy in patients with multiple endocrine neoplasia type 2 or colectomy in patients with familial adenomatous polyposis. Again, for more prevalent disorders and susceptibility markers, markers must be prevalent enough to adequately power outcomes-based clinical trials.

Finally, for genetic markers of susceptibility to gain acceptance into clinical practice, the costs of testing must be manageable. With efforts such as the personal genome project^[42] and rapid advances in sequencing technology, the cost of screening will likely become more affordable. However, reimbursement strategies for genomic testing must still be established. Thus, although there is no question that genomic data are pointing toward novel pathways and mechanisms underlying complex diseases, it remains to be seen whether these data will also translate into useful clinical recommendations.

Pharmacogenetics and Pharmacogenomics

Optimizing Drug Response

The impact of the genome on our ability to predict drug response is one of the most promising and fertile areas of genomic and personalized medicine. Pharmacogenetics is the study of genetic variation that ultimately gives rise to the variable responses in individuals to any given drug treatment. More recently, pharmacogenetics has provided an explanation as to why certain individuals do not respond, or respond differently, to a given drug treatment. Pharmacogenomics uses genomic technology to understand the effects of all relevant genes on the behavior of a drug or conversely the effect of a drug on gene expression. Pharmacogenomics, like pharmacogenetics, has rapidly embraced genomic technologies to identify molecular patterns of response, drug disposition, and drug targets. Both approaches have great potential to positively affect the field of medicine.

Pharmacogenomic Tests: Promise on the Horizon

Perhaps the best example of a successful pharmacogenetic association for which the clinical relevance is clear is the management of warfarin therapy.^[43] The oral anticoagulant warfarin (a derivative of coumarin) is commonly prescribed for the long-term treatment and prevention of thromboembolic events. However, because of the drug's narrow therapeutic index, a variety of complications are associated with its treatment, even after dose adjustment according to age, gender, weight, disease state, diet, and concomitant medications. Investigation of pharmacokinetic and pharmacodynamic drug properties indicated the additive involvement of 2 genes in determination of warfarin maintenance dose. One of these genes encodes CYP2C9, which is responsible for most of the metabolic clearance of the more pharmacologically potent S-enantiomer of warfarin. Both CYP2C9*2 and *3 cause a reduction in S-warfarin clearance, with the lowest activity variant (CYP2C9*3) showing 3% to 11% of the activity of the most active (wild type) variant, CYP2C9*1. Numerous studies have associated these genotypes with initial dose sensitivity, delayed stabilization of maintenance dose, delays in hospital discharge and increased bleeding complications.^[44] However, it is estimated that CYP2C9 variants account for only 6% to 10 % of the total variation in warfarin dose,^[45] with additional genetic and environmental factors playing larger roles in dose determination. The second gene identified as a predictor of dosing is the vitamin K epoxide reductase complex protein 1 (VKORC1), targeted by warfarin and accounting for 21% to 25% of dosage variance.^[45] According to new product labeling, consideration of VKORC1 genotype or haplotype together with CYP2C9 genotype and factors such as age and body size are estimated to account for about 55% of the variability in warfarin dosing requirements.

Despite the fact that multiple independent groups have reproduced these data, prospective clinical studies are required to establish whether initial dose may be tailored to patients by CYP2C9 and VKORC1 genotyping coupled with known clinical variables.^[46] These studies are currently underway, however, the Food and Drug Administration acknowledged the importance and potential for genotyping of CYP2C9 and VKORC1 during the early phase of warfarin therapy, and the drug label was amended accordingly in August 2007.

More than 50 years ago, 6-mercaptopurine was marketed for the treatment of acute lymphoblastic leukemia. Despite great expectations, fatal bone marrow suppression was found in 0.3% of treated children. There were similar findings for azathioprine several years later. It was later discovered that polymorphisms within the thiopurine methyltransferase (TPMT) gene underlie the large interindividual differences in the enzyme's activity, leading to a high risk for thiopurine-induced toxicity in homozygotes for the defective alleles and inadequate therapeutic efficacy in patients with high-activity TPMT.^[47] Tests for TPMT activity (genotype, enzyme activity, and metabolite screening) are available in the United States and throughout Europe; however, clinical implementation of these tests is very low. This is contrary to expectations based on cost-effectiveness analysis of TPMT testing in children with acute lymphoblastic leukemia showing high savings per life-year.^[48] Current clinical practice for the management of leukemia using thiopurines dictates careful monitoring of white blood cell counts and clinical outcomes. It is expected, however, that as genetic tests become generally accepted for a variety of conditions, it will become progressively acceptable to use TPMT genetic testing as a prognostic tool for adverse drug response.

A forme fruste of pharmacogenomics is the notion of targeted therapies.^[49] Trastuzumab therapy (a monoclonal antibody specifically targeting HER2/neu-overexpressing breast tumors) for the treatment of breast cancer is an example of a protein therapeutic for which an obligatory biomarker assay and diagnostic test has been developed to identify the patients most likely to benefit from this drug. Trastuzumab is marketed solely for the subset of patients who have overexpression of HER2/neu.Given the low prevalence of marker-positive breast cancers, it is conceivable that if it were not for the use of the diagnostic marker in clinical development, the drug would not have been successfully developed. Cancer is not the only field of medicine with a targeted pharmacogenomic approach to giving therapeutics. In cardiovascular medicine, a targeted approach to acute coronary syndromes has been practiced for more than a decade with the use of cTnI measurements to dictate the beneficial use of glycoprotein IIb/IIIa inhibitors.

To assist clinicians in the practice of pharmacogenetics across a broad number of medications, the first microarray-based gene chip, approved both in the United States and Europe, was released in 2003 as the AmpliChip CYP450.^[50] The product was designed to identify key genetic polymorphisms in 2 CYP450 enzymes, CYP2D6 and CYP2C19, cumulatively responsible for much of the first-pass metabolism of many currently prescribed drugs. The regulatory agencies indicated that its utility as a stand-alone test remained unproven.^[51] Thus clinicians, as well as patients, are unclear about the impact of these tests on clinical decision-making guidelines. Until unambiguous evidence proves the clinical use of this and other genetic tests, caution is advised in their interpretation and application in healthcare management.

Can interpatient variability in somatic tissues such as a tumor be used in treatment planning? Recently, a series of gene expression signatures have been developed that predict response and resistance to conventional cytotoxic chemotherapeutic agents, portending the advent of personalized cancer treatment based on a tumor's gene expression pattern. Using in vitro drug sensitivity data combined with microarray gene expression data publicly available for the NCI-60 set of cell lines, signatures of response to , docetaxel, paclitaxel, topotecan, doxorubicin, cyclophosphamide, and etoposide were developed using logistic regression modeling.^[52] Using independent sets of human tumors with known clinical outcome and for which expression data were available, the authors went on to demonstrate that these signatures could predict response to chemotherapy in these tumors in both the neoadjuvant and adjuvant chemotherapy settings. These genomic signatures now form the basis for a series of first-of-their-kind clinical studies in which treatment assignments in the trials are being made based on the pharmacogenomic molecular signatures from a patient's tumor. This is one of the clearest examples of a genomic technology paving the way for truly personalized medical treatment.

Getting Personal

The Personal Genome

The advances in technology in sequencing make the concept of having a personal copy of one's genome quite realistic in the next decade. Today, it will cost about $350,000^[53] and takes several months. The X-prize initiative offers $10 million to the first team to sequence 100 human genomes in 10 days.^[54] If the National Institutes of Health grant initiatives are successful, it will cost closer to $1000.^[55] In anticipation of the disruption in medical care this approach may cause, the Personal Genome Project was launched with the potential aim to publish the complete genomes and medical records of volunteers to enable research into personalized medicine.^[42] All data will be freely available over the Internet, so that researchers can test various hypotheses about the relationships among genotype, environment, and phenotype. The long-term goal is for every person to have access to his or her genotype to be used for personalized medical decisions.

Today, advances in SNP technology are allowing a glimpse of the variation in our genomes at a much lower cost. Three companies (Navigenics, 23andMe, and DeCode) have announced direct-to-consumer initiatives to make variation in one's genome available for $1000 to $2500. These initiatives use the 500K to 1000K SNP chip or similar technologies. They report information back to their customers on 20 to 30 areas of the genome that have been identified as disease-risk susceptibility loci and some offer additional information such as ancestry. How this shift in healthcare information delivery will integrate into medical practice remains to be seen. It will not be long before a patient will bring this report to your office and ask for guidance. What will you tell him or her? Not only will you need to know the importance and clinical implications of the information, but you will also need to know how to integrate genetic risk information with clinical risk information and effectively communicate that risk to your patient. The science of risk communication to affect behaviors is a science that is evolving along with the genomic technologies that are unraveling that risk.

Whether SNPs are detected or sequencing is done, our genomic information will be available in a short period of time. A clear and important research agenda needs to be developed in concert with these technologic breakthroughs that allows health providers and the public to understand the information and, more important, to believe that it is accurate, informative, and actionable. Mark Twain has been quoted as saying "Beware of health books. You might die of a misprint.^[56]" With the vast amount of information contained in the human genome sequence, the stakes are high for all of us to ensure that the proper reading and interpretation of the information are carried out.

Integration of Genomic Testing Into Clinical Practice

Barriers

Despite the optimism about what genomic testing might do for medicine, there are barriers that must be overcome before it can be integrated into clinical practice. The incorporation of genetics and genomics into patient management guidelines has largely failed to occur, perhaps because researchers, diagnostic firms, and the regulatory authorities are still seeking to establish methodologies by which to judge their effectiveness, practicing clinicians and guideline writers are still working to understand how such new tests fit into current models of care and risk assessment, and payers are just beginning to foresee new pressures to cover the additional costs. We have proposed a framework to assist in genetic testing evaluation.^[57]

Summary and Future Directions

The undeniable allure of genome technology as applied to medicine is high. But what of the realities of bringing genomic medicine to the clinic and seamlessly integrating it within current models of healthcare delivery? This is where the steepest barriers lie. Although the human genome sequence is now available, it is important to acknowledge that our knowledge of the genome and its biological complexity is nowhere near complete, and the installation of genomic protocols into standard clinical care is virtually unknown territory. There are a host of clinical, economic, insurance, privacy, and commercialization concerns that will need to be addressed and that vary substantially among different countries. And of course, before we can confront those, we must be certain that genomic medicine is on the soundest possible scientific footing. If those issues can be dealt with systematically, the prospect of using genomic information to offer patients healthcare that is truly prospective in nature may finally be within our grasp.

This activity is supported by an independent educational grant from Navigenics.