NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Institute of Medicine (US) Forum on Neuroscience and Nervous System Disorders. Neuroscience Biomarkers and Biosignatures: Converging Technologies, Emerging Partnerships, Workshop Summary. Washington (DC): National Academies Press (US); 2008.

Cover of Neuroscience Biomarkers and Biosignatures

Neuroscience Biomarkers and Biosignatures: Converging Technologies, Emerging Partnerships, Workshop Summary.

Show details

2Potential Tools for Biomarker and Biosignature Development

Biomarker development hinges on the effective advancement and resourcing of currently available tools and technologies. Outcome measures include treatment efficacy, increased specificity in clinical trials and therapeutics, and identification of target study molecules. In Session II, workshop participants discussed the value of genomics, proteomics, and imaging as tools for bio-marker discovery and development.


Genome-wide scanning is a relatively new genetic technology for finding biomarkers associated with disease. It is a method of scanning the entire genome in the search for single nucleotide polymorphisms (SNPs) that are correlated with disease. While the vast majority of SNPs are innocuous, SNPs associated with disease are identifiable in combination with other data, including epidemiology studies that compare large groups of individuals with the disease against other groups without the disease. When SNPs associated with disease are found on particular regions of the chromosome, these SNPs subsequently can be used to pinpoint disease-specific loci to the disease-related gene.

An outgrowth of the Human Genome Project and the International HapMap Project, genome-wide scanning has a myriad of applications, including the identification of targets for drug development. Dr. Allen Roses, senior vice president of GlaxoSmithKline, Inc., focused his presentation on genomics’ impact for development of biomarkers for nervous system disorders. He pointed out that genotypes associated with disease eventually may be used to predict which patient groups are more susceptible to disease, which are more likely to experience adverse effects of drugs, or which are more likely to benefit from drug therapy and at what doses (among other applications).

Genome-wide scanning already has been applied successfully to at least one nervous system disorder, Alzheimer’s disease (Martin et al., 2000), and is soon expected to yield results for schizophrenia, according to Roses. It is now well established that the gene APOE is a susceptibility gene for Alzheimer’s disease (Roses, 1996). Drawing from this pioneering work, Roses first focused on the value and efficiency of genome-wide scanning as a method to validate and confirm genetic loci first found by previous methods that were more labor intensive (Lai et al., 1998). Genome-wide scanning, said Roses, has an equally important role in disconfirming other loci identified by earlier methods. Narrowing the search for the most important loci is essential before undertaking the laborious process of finer mapping and positional cloning to find, within the loci, specific genes that are defective. Genome-wide scanning also can be used, on its own, without being hypothesis-driven about which chromosomal regions to search. In other words, it can be used in a hypothesis-free manner to examine new regions of the genome not explored in earlier studies. Those studies were often small, family-based association studies rather than the large population cohorts now being studied in association with genome-wide scanning.

With genome-wide scanning, around 500,000 SNPs are now used to examine the entire human genome to identify possible target loci for disease biomarkers. This figure represents a small subset of the 10 million SNPs found across the genome. The reason it is possible to examine only a small subset of SNPs is because of the correlation (linkage disequilibrium) that exists between SNPs in close proximity to one another. The 500,000 SNPs now available by commercial microarray technology capture 80 percent to 85 percent of the entire genome (and these figures are now growing with newer versions of the technology). After conducting the first analysis to find SNPs associated with disease, a series of replication analyses are performed with the same or with larger cohorts to eliminate false positives (considering the huge number of comparisons being made in a large cohort, as opposed to smaller, family-based associated studies, which are less sensitive to false positives). A combination of methods, for example, was used to implicate another gene associated with APOE. An analysis that focused on the gene SORL1 found no variants involved in defective processing of amyloid precursor protein in the pathophysiology of Alzheimer’s disease (Rogaeva et al., 2007).

Roses also provided examples of the value of genotype biomarkers in clinical trial design. Results from clinical trials can be highly dependent on the genotype of the patient (Roses et al., 2007). Therefore, enriching trials with patients who have the receptive genotype is expected to enhance the likelihood of demonstrating drug efficacy and reduce the size of the trial. One example occurred during a trial of rosiglitazone, a drug targeted to combat the APOE defect in Alzheimer’s disease. Overall, the combined group of patients with mild and moderate Alzheimer’s disease did not improve with the drug, but after the patients were stratified by genotype, it was recognized that APOE4-negative patients improved with the drug, whereas those who were APOE4-positive failed to improve (Risner et al., 2006). Because the original data had been pooled together, the drug program would have been needlessly halted from lack of efficacy. Once the value of genotyping was established, subsequent phases of the clinical trial were redesigned and powered appropriately to ensure that the drug’s effects would be realized among patients with the susceptible genotype. Three genotype-specific Phase III clinical trials are in progress for rosiglitazone, noted Roses, with a subset of patients carefully selected by genotype. Stratification by genotype also was crucial for choices about dose. Higher doses of the drug were needed to see a positive effect in APOE4-negative patients. The concept of sequential analysis in clinical trial design—using each phase to help enrich subsequent phases with genotype-specific patients—has broad applications for drug development. The one major concern, however, is that the genotype being targeted by genome-wide scanning (or other methods) may be too specific in its physiological effect and, thereby, miss other candidate genes with broader therapeutic effects.


Proteomics-based biomarker discovery is a relatively new field that can be harnessed to identify new central nervous system (CNS) bio-markers. In the broadest terms, proteomics seeks to understand the total protein complement in fluids or tissues by identifying individual or groups of proteins, their levels of expression, post-translational modification, and protein-protein interactions, among other characteristics from which protein and cellular function can be inferred. The field, however, does not yet have the capacity to conduct a whole-proteome scan to the same extent that the entire genome can now be scanned; proteomics can, however, identify hundreds to thousands of proteins in small samples of complex fluids or tissues. The proteome, unlike the genome, differs from cell to cell and, over time, changes dynamically within each cell in response to external stimuli. Its ability to report on the physiological state of the organism is what makes it valuable as a source of biomarkers but is also what makes proteomics more challenging.

Against this fluctuating background in protein expression, the search for biomarkers and biosignatures of disease looks for reproducible changes expressly associated with disease or response to drugs. The study of CNS diseases, while in its infancy, will be greatly aided by identifying patterns of expression of multiple protein biomarkers in the same way that the measurements of HDL, LDL, and cholesterol are biosignatures of cardiac disease, noted Dr. Howard Schulman, vice president of R&D of PPD Biomarker Discovery Sciences.

Identification of potentially new protein biomarkers of disease first requires extraction of fluids or tissues, protein separation, identification, and quantification of relative levels of protein expression using advanced software. The most advanced methods of protein profiling rely on liquid chromatography combined with mass spectrometry (LC/MS), said Schulman. In his presentation, he described PPD’s main approach to proteomic discovery as “an unbiased screen of several thousand potential biomarkers in biological fluids or tissues.” The workhorse of proteomics historically has been the use of two-dimensional gels to separate proteins, but this method has lower sensitivity and throughput than LC/MS. These approaches compare with hypothesis-based approaches, using more narrowly targeted methods that typically use antibody reagents, for example, to find ratios in the levels of a small number of proteins. But the lack of antibody reagents has been a rate-limiting problem in the application of proteomics to the CNS. Hypothesis-based approaches or multiplexed screens with panels of antibodies can complement proteomic discovery by measuring low-abundance proteins. In general, the proteomics field is less developed for applications to the CNS than for other bodily systems, largely because fluids and tissues from the CNS are less accessible.

LC/MS generates relative concentrations of proteins by measuring signal intensity, but it cannot generate absolute concentrations. A major hindrance to protein profiling is the broad dynamic range of protein concentrations found in a complex mixture. To overcome this problem, one approach begins by affinity removal of the 80 percent to 90 percent of the protein mass contributed by the 6 to 12 most abundant proteins. Further fractionation of the mixture can yield certain proteome classes, such as low-molecular-weight (peptidome) versus higher-molecular-weight proteins (Figure 2-1). Peptidomes are often 10 times less abundant in the sample. Another approach is to subdivide proteins by attached chemical group, such as phosphoproteins, glycoproteins, and ubiquitinated proteins (i.e., the key to dealing with the problem of a wide dynamic range is through subsampling and fractionation).

FIGURE 2-1. Differential quantification: proteins and peptides.


Differential quantification: proteins and peptides. NOTE: Molecular weight (MW); high-performance liquid chromatography (HPLC); electrospray ionization mass spectrometry (ESI-MS); mass spectrometry (MS); combination of two or more MS experiments (MS/MS). (more...)

For the purpose of discovery in CNS diseases, proteomics is best accomplished by examining samples from the cerebrospinal fluid (CSF) rather than from the blood, Schulman stressed. The CSF carries higher concentrations of biomarkers because it is closer to the source of the pathology and the physiological response to it (Huhmer et al., 2006). CSF is enriched with intracellular proteins and proteins in extracellular debris that are likely associated with disease (Schulman, 2006). The CSF also has a smaller dynamic range of protein concentrations to facilitate in-depth analyses, meaning that a smaller range exists between the most and the least abundant proteins. Its dynamic range is an order of magnitude less than that of blood. From about 1\of CSF it is possible to profile 1,000 to 2,000 proteins.

Another obstacle is public attitudes, which view lumbar puncture as too invasive. This view is misplaced as long as the potential for benefit is strong, said Schulman, who notes that lumbar puncture is well accepted in Europe and Scandinavia. The evidence from one of the first CNS diseases for which biomarkers are being developed, Alzheimer’s disease, justifies a shift in American attitudes. Proteins directly associated with the disease are detectable in CSF but are only poorly detectable in the blood, or their levels do not reflect changes in the CNS levels of the protein (Irizarry, 2004). A shift in American attitudes is likely to occur once more is known about the low risks associated with lumbar puncture. Schulman pointed out that in Potter’s presentation, for example, attitude toward lumbar puncture was suggested to be improved as a result of subjects viewing an educational video that profiled the low risks associated with lumbar punctures.

In the case of CNS lymphoma, lumbar punctures are routinely done for cytological tests even though the results are not diagnostic, with lymphoma cells only detected in about 40 percent of the subjects with the cancer. Schulman reported that he and his colleagues are developing better methods to diagnose CNS lymphoma with CSF, first by seeking to correlate potential biomarkers from CSF with imaging studies and other clinical indicators of disease. The initial set of biomarkers is already better than the existing cytological test (Rubenstein, 2005). Biomarkers found in the spinal cord have the potential not only as a diagnostic test but also as a measure of response to treatment. It is possible that after initial discovery of a useful biomarker from the CSF, a blood test can be created to check for that particular biomarker.

Apart from Alzheimer’s disease, for which biomarkers are being developed, proteomics is only beginning to be investigated for diseases arising within the CNS. Some of the most obvious diseases for study include schizophrenia, depression, and autism (Box 2-1). Some of the major challenges in further expansion of the field are the limited number of reagents for enriching and subsampling classes of proteins, the wide dynamic range of concentrations of brain proteins (which means that many proteins would be missed), and improving the sensitivity of the LC/MS. To narrow the search, new methods need to be developed to eliminate the most abundant peptides. Reagents for depleting abundant proteins are typically designed for plasma proteins, but proteins in the CSF do not completely overlap with them. Finally, overcoming the public’s attitude toward the invasiveness of lumbar puncture is key to CNS biomarker development with proteomics.

Box Icon

BOX 2-1

Biomarker Opportunities in Neuroscience. Comprehensive phenotyping (discovery- and hypothesis-based approches; protein, genes, imaging, etc.) includes Antecedent markers in schizophrenia


Imaging not only occupies a singular place in the current practice of medicine, but also holds enormous prospects for future biomarker development, according to Dr. Bruce Rosen, director of the Center for Biomedical Imaging at Massachusetts General Hospital. Rosen opened by citing a survey of practicing internists who selected computed tomography (CT) and magnetic resonance imaging (MRI) as the leading medical advances of the past quarter-century. They were ranked first, superceding 30 other possible advances, including ACE inhibitors, statins, and mammography (Fuchs and Sox, 2001). Rosen profiled some existing biomarker advances, numerous cutting-edge opportunities, and several key barriers to progress for the two main types of imaging biomarkers—structural and functional. In years to come, both are likely to be integrated together in often creative and revealing ways.

The most common structural biomarkers being applied to the nervous system include CT and MRI, and functional biomarkers include positron emission tomography (PET; for neurochemistry), electrophysiology by electroencephalography and magnetoencephalography (MEG),1 and functional MRI (fMRI); some of the imaging outcomes are described later in this section. Their ultimate value, from Rosen’s perspective, is to provide surrogate markers for eventual qualification by the Food and Drug Administration. The markers they generate might represent early pathophysiologic indicators of disease, diagnosis, or treatment (especially dosing and response to treatment). Rosen noted that one of the major successes of the imaging field, which comes from oncology, is a structural biomarker showing tumor volume reduction in not only one but several colon cancer clinical trials. The reduced size of the tumor showed a strong correlation with overall survival and thereby could be developed as a surrogate to hasten the pace of drug development (Fleming, 2005). But for every success, there are failures. One guidepost for finding a successful biomarker, said Rosen, is to focus on those that participate in the pathophysiological process of the disease under study.

Rosen summarized the status of biomarker development in neuro-imaging, emphasizing that the best anatomical biomarkers have been in quantitative morphometry and white matter conductivity. The greatest functional biomarkers have been in several areas of physiology, metabolism, receptor distribution, and electrophysiology (Box 2-2).

Box Icon

BOX 2-2

Neuroimaging as an Indicator for Neural States. Anatomy Quantitative morphometry

In the near future Rosen anticipates the most progress in utilizing gray matter ultrastructure (which is especially important for developmental diseases), high-resolution cytoarchitectonic mapping, tomographic electrophysiology (which will enable identification of functional conductivity patterns), and the combination of imaging biomarkers with markers of gene and protein expression. As tantalizing as these prospects may be, Rosen anticipates that structural markers will continue to be a mainstay for a long time to come.

As one of many examples of the cutting-edge utility of imaging, Rosen highlighted some recent developments in Alzheimer’s disease biomarkers. For years there has been debate about whether regional thinning in the brain of Alzheimer’s patients reflects the disease process versus normal aging. New research reveals that Alzheimer’s disease is indeed distinguishable from normal aging by distinct patterns of thinning in the parietal cortex, posterior hippocampus, and entorhinal cortex—nuclei long associated with Alzheimer’s disease. These and related accomplishments are made possible by more sophisticated ways to increase resolution, allowing visualization of individual nuclei or tracts within the brain (Figure 2-2). The entorhinal cortex, for example, consists of nests of about 100 cells, each of which are about 250 microns in size, said Rosen, who noted that this degree of resolution has not been heretofore achieved. This nucleus is highly important because it is the first group of cells to die during early Alzheimer’s disease. A related example is from a study showing that in response to a cognitive task, people with attention deficit hyperactivity disorder (ADHD) fail to activate their anterior cingulate nucleus in comparison with normal controls (Bush et al., 1999).

FIGURE 2-2. Selective regional thinning in Alzheimer’s disease.


Selective regional thinning in Alzheimer’s disease. SOURCE: Adapted from Salat et al., 2004.

The more distant prospects for neuroimaging are wide ranging. They include circuitry-based diagnosis of such disorders as ADHD, substance dependence, depression, schizophrenia, and obsessive-compulsive disorder; better contrast agents to increase sensitivity to visualize blood volume; imaging of gene expression through MR; and molecular imaging of substance abuse (Volkow et al., 2006) (or other brain-based disorders through PET scanning or pharmacologic MRI2).

An attractive goal, highlighted by Rosen, is to develop combination modalities that integrate structural and functional information, such as anatomical MRI in combination with functional and baseline perfusion MRI. In one early example, investigators compared five different modalities for imaging the hippocampus in Alzheimer’s disease. The modalities are MRI, PIB (a radiotracer for an amyloid ligand), FDG (a radiotracer for glucose metabolism), ASL (arterial spin labeling MRI that measures blood perfusion), and fMRI. The combination of PET and MRI might enable study of receptor-specific functional activation through simultaneous physiology and receptor kinetics. The combined approach ultimately may add electrophysiology, predicted Rosen, who said that several modalities combined, rather than a single modality, may become the best markers for profiling nervous system diseases.

None of the opportunities are without barriers to development, the most common of which are

  • the need for validation of images to ensure face validity (among other forms of validity testing);
  • standardization of imaging protocols across centers, particularly as scanners are upgraded in field strength;
  • the need for sophisticated informatics to integrate the information provided by markers from several different imaging modalities, in addition to levels of gene and protein expression;
  • the formidable blood-brain barrier across which drugs or radio-tracers must penetrate; and
  • the insufficient number of radiotracers for PET studies that are tailored to proposed molecular defects. This problem was echoed by Dr. Nora Volkow, director of the National Institute on Drug Abuse, who pointedly called the lack of tracers the “strongest impediment to progress.”

The absence of radiotracers has become such a rate-limiting problem that a new public-private initiative has been proposed to tackle it. The Radiotracer Clearinghouse (RCH) is a nonprofit organization providing a solution to help fast-track drug discovery and development processes for CNS and any other therapeutic areas. RCH was conceived as a vehicle to enable the pharmaceutical industry to share information on radiotracers within a secure environment designed to protect all parties’ intellectual property. Within the RCH, scenarios may include sharing information (1) under strict confidence with minimum disclosure between parties and no public disclosure or (2) with all parties involved with intent to publicly disclose information related to the biomarker, target, or specific imaging study in a timely manner. The latter scenario may be covered under RCH as a project with the Biomarker Consortium. In all cases, the rules of engagement for the scope and timing of information shared throughout the process will be established before each project begins by the RCH facilitator and the pharmaceutical companies/academic partner involved, according to Dr. Dean F. Wong, professor of radiology and psychiatry, vice chair radiology research and section director of high resolution brain imaging at the Johns Hopkins University.

MEG measures magnetic fields produced by electrical activity in the brain

Pharmacologic MR uses pharmacological challenge to image neurocircuits and to track drug time course through hemodynamics.



MEG measures magnetic fields produced by electrical activity in the brain


Pharmacologic MR uses pharmacological challenge to image neurocircuits and to track drug time course through hemodynamics.

Copyright © 2008, National Academy of Sciences.
Bookshelf ID: NBK53108
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.5M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...