| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© Oxford University Press 2007.
NEWS |
SNPs Not Living Up to Promise; Experts Suggest New Approach to Disease ID
In June 2000, researchers announced to great fanfare that a rough draft of the human genome had been decoded. That achievement—followed up with a completed draft 2 years later—was heralded as the gateway to personalized medicine, a new paradigm for treating patients according to their genetic makeup.
Writing in the New England Journal of Medicine, Francis Collins, M.D., Ph.D., predicted that gene-based primary care would be thriving by 2010. "This knowledge will dramatically accelerate the development of new strategies for the diagnosis, prevention, and treatment of disease," proposed Collins, who directs the National Human Genome Research Institute (NHGRI), in Bethesda, Md., " ... not just for single-gene disorders but [also] for the host of more common, complex diseases [e.g., cancer]."
But today, researchers are substantially downplaying those prospects. "We have more modest, but realistic, expectations," says John Ioannidis, M.D., professor and chair of the University of Ioannina School of Medicine in Greece. He thinks that they will be able to get to individualized medicine but that the steps will be more incremental than monumental.
|
Despite much research, there has been limited success in using the human genome to advance personalized care. Genetic contributions to disease are much more complex than originally anticipated. The human genome has produced a great many complex and sometimes ambiguous findings that don't yield clear opportunities for treatment yet.
Ioannidis says personalized medicine's slow pace stems from the unexpected challenge of linking diseases to specific genetic variations, especially single-nucleotide polymorphisms, or SNPs. "Most of us thought it would be easier to associate SNPs with disease," he says. "But the proposed associations haven't held true upon further testing."
SNPs are clinically important because they can amplify disease risks and offer new targets for drugs. For example, the risk of ovarian and breast cancer increase up to 85% and 65%, respectively, for women who harbor a SNP called IVS8+2T>A. That SNP substitutes adenine for thymine in the gene that codes for BRCA1, a protein that normally repairs damaged DNA. When compromised by the sequence change, BRCA1 can't perform its natural role and the cancer risk increases.
Women with variations in BRCA1 (and those in a related gene called BRCA2) receive the long-envisioned personalized care that can include heightened surveillance, chemoprevention, and prophylactic surgery. With the completion of the human genome project, researchers had expected that many similarly predictive SNPs would soon be discovered, which would greatly expand personalized medical care.
But so far, that hasn't been the case: Despite 25,000 research papers investigating disease associations with SNPs in nearly 2,500 genes since 2000, researchers have been unable to find many new variations with substantial clinical effects, according to Muin Khoury, M.D., Ph.D., who directs the National Office of Public Health Genomics at the Centers for Disease Control and Prevention. At most, 10% of published associations have been validated by other researchers, Khoury says. And of these, few if any have resulted in clinical or public-health applications.
Khoury attributes the many SNPs in the literature to "publication bias," the tendency among scientific journals to publish positive results for newly described SNPs even if detected in underpowered studies with few subjects. Other research groups then rush to validate those findings, he says, which partly explains why thousands of research articles about SNPs appear every year. This trend represents a substantial investment in funding resources.
Even validated associations often crumble when subjected to large-scale reviews, Ioannidis adds. He cites a recent meta-analysis by the Breast Cancer Association Consortium, a collaboration of more than 20 research groups with a combined sample size exceeding 30,000 patients and 30,000 controls. The BCAC study (J Natl Cancer Inst 2006;98:1382–96) focused on SNPs whose association with breast cancer had been studied by at least three separate laboratories. Of the 16 considered, just five achieved borderline statistical significance, whereas the rest were not linked with breast cancer at all. In an accompanying editorial, Ioannidis wrote that 16 more SNPs previously linked to breast cancer had also failed in a separate meta-analysis. That finding underscores what he says is a "sobering picture."
"We've got a vast number of hypotheses being tested, but most of them don't get anywhere. When you really look at them, the associations disappear," he says.
An Environmental Perspective
Responding to these failures, some researchers have revised their notion of how most SNPs participate in disease and how they might guide care. "We no longer expect to find SNPs with big effects," says Regina Santella, Ph.D., a professor of environmental health sciences at Columbia University's Mailman School of Public Health. "Most of the major SNPs have already been found. We don't think there are many more out there. The effects of the remaining SNPs are probably small and interactive. It's not just one or two SNPs that have an impact on health; it's likely to be multiple genotypes interacting with each other, and with environmental factors."
Indeed, this conclusion may bring the environment's influence on SNP contributions to the forefront, experts say. Christine Ambrosone, Ph.D., chair of the department of cancer prevention and control at the Roswell Park Cancer Institute, suggests that the BCAC study might have detected associations with cancer if environmental factors—such as exercise, diet, tobacco use, pollutants, and industrial chemicals—had been incorporated in the analysis. For instance, a SNP called ADHIC, which is among the 16 studied by the BCAC, might exacerbate cancer risk only for heavy alcohol drinkers, she says. For its part, the BCAC members write that neither gene-to-gene nor environmental effects were considered, because doing so would have increased the risk of false-positive results even further.
But it's not just statistical challenges that complicate studies of SNP–environment interactions. The investigations also suffer from a chronic shortage of environmental exposure data and a lack of specificity in the data they have. Today, researchers typically estimate current and past exposures from questionnaires, which are prone to bias and other shortcomings. Actual biomarkers that reflect how, when, and where environmental agents interact with DNA would be more useful, but such biomarkers are rare, according to John Groopman, Ph.D., chairman of the department of environmental health sciences at the Johns Hopkins Bloomberg School of Public Health. That's particularly true for biomarkers of past exposure, he says, which affect cancers that appear later in life.
NHGRI and the National Institute of Environmental Health Science (NIEHS) have recently begun a collaboration focused on gene–environment interactions. As part of this effort, NIEHS scientists will attempt to produce new exposure biomarkers for use in genetic research. But just as SNPs require validation to be clinically useful, biomarkers too must be validated, ideally in large cohorts subjected to repeated sampling. These studies are time consuming and expensive, so in all likelihood new biomarkers will become available slowly.
No More Candidate Genes
In the meantime, genetic research has been shifting away from a traditional approach to SNP discovery. Early on, researchers looked for SNPs primarily within "candidate genes" emerging from pathway-directed studies in molecular biology. But that method hasn't been fruitful, Ioannidis says, because pathway-focused studies often point in the wrong direction. Researchers typically don't know enough about the underlying genetic mechanisms in disease, he explains, so they don't choose candidate genes correctly. "We put too much trust in the little biology we know," he says. "We go after candidate genes based on weak evidence."
In a shifting strategy, researchers are using whole-genome scanning to search for disease genes. This technique casts a wide net over the genome, allowing researchers to evaluate thousands of genes simultaneously without prior assumptions about underlying disease mechanisms. Falling costs have allowed this more unbiased technique to flourish; researchers say it could be better suited to investigations of small genes with small effects. However, the issue of false-positive findings remains a challenge.
Kathy Helzlsouer, M.D., who directs the center for prevention and research at Mercy Medical Center in Baltimore, Md., cautions that although whole-genome scanning might reveal new disease pathways, it still has key limitations. Genes make proteins that often change in key ways after transcription, she says. And the influence of those changes on cancer risk can't be detected by focusing exclusively on the genome.
"There's lots going on at the posttranslational side," Helzlsouer explains. "And that's where genetic side has a disadvantage; it doesn't look at this critical element." Nevertheless, Helzlsouer suggests that the combination of candidate gene and genomewide approaches could produce synergistic benefits. "We still have to find out which genes are important," she says.
To turn the research around, Ioannidis also suggests that "harmonized" efforts involving multiple research groups and an integrated effort could make SNP discovery more efficient and effective. Researchers need large cohorts to detect SNPs with small effects, he says. But large cohorts stretch the administrative capacity of individual laboratories, which underscores the need for collaboration. Ioannidis cautions that even broader research networks might be unable to detect rare gene variations that confer minor risks. "Even with these advances, we might still be stuck," he says.
Ultimately, the opportunities that SNPs might provide for medical care remain unknown. The experts interviewed emphasize that personalized medicine may have been overhyped during its introduction to the public. "I would agree that if we're really looking at hundreds of thousands of SNPs with small effects, then they may not be clinically useful," said Mary V. Relling, Pharm.D., chair of the pharmaceutical department at St. Jude Children's Research Hospital. "I'm not willing to say that's the final conclusion; it's a depressing notion, but I'm not there yet."
Khoury adds, "I think in 10 years we'll know if we've just created a data monster or if we've pushed the science forward. Although I reserve final judgment, I tend to be optimistic. I think we'll learn more if we apply these tools judiciously."
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  P. A. Thompson Counterpoint: Genetic Risk Feedback for Common Disease Time to Test the Waters Cancer Epidemiol. Biomarkers Prev., September 1, 2007; 16(9): 1727 - 1729. [Full Text] [PDF]
|
|
|
|
|
|
P. M. Webb, M. A. Merritt, G. M. Boyle, and A. C. Green Microarrays and Epidemiology: Not the Beginning of the End but the End of the Beginning... Cancer Epidemiol. Biomarkers Prev., April 1, 2007; 16(4): 637 - 638. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||