may have to wait. The genetic analysis of common disease is turning out to be a lot more complex than expected.
Since the human genome was decoded in 2003, researchers have been developing a powerful method for comparing the genomes of patients and healthy people, with the hope of pinpointing the DNA changes responsible for common diseases.
This method, called a genomewide association study, has proved technically successful despite many skeptics’ initial doubts. But it has been disappointing in that the kind of genetic variation it detects has turned out to explain surprisingly little of the genetic links to most diseases.
A set of commentaries in this week’s issue of The New England Journal of Medicine appears to be the first public attempt by scientists to make sense of this puzzling result.
One issue of debate among researchers is whether, despite the prospect of diminishing returns, to continue with the genomewide studies, which cost many millions of dollars apiece, or switch to a new approach like decoding the entire genomes of individual patients.
The unexpected impasse also affects companies that offer personal genomic information and that had assumed they could inform customers of their genetic risk for common diseases, based on researchers’ discoveries.
These companies are probably not performing any useful service at present, said David B. Goldstein, a Duke University geneticist who wrote one of the commentaries appearing in the journal.
“With only a few exceptions, what the genomics companies are doing right now is recreational genomics,” Dr. Goldstein said in an interview. “The information has little or in many cases no clinical relevance.”
Unlike the rare diseases caused by a change affecting only one gene, common diseases like cancer and diabetes are caused by a set of several genetic variations in each person. Since these common diseases generally strike later in life, after people have had children, the theory has been that natural selection is powerless to weed them out.
The problem addressed in the commentaries is that these diseases were expected to be promoted by genetic variations that are common in the population. More than 100 genomewide association studies, often involving thousands of patients in several countries, have now been completed for many diseases, and some common variants have been found. But in almost all cases they carry only a modest risk for the disease. Most of the genetic link to disease remains unexplained.
Dr. Goldstein argues that the genetic burden of common diseases must be mostly carried by large numbers of rare variants. In this theory, schizophrenia, say, would be caused by combinations of 1,000 rare genetic variants, not of 10 common genetic variants.
This would be bleak news for those who argue that the common variants detected so far, even if they explain only a small percentage of the risk, will nonetheless identify the biological pathways through which a disease emerges, and hence point to drugs that may correct the errant pathways. If hundreds of rare variants are involved in a disease, they may implicate too much of the body’s biochemistry to be useful.
“In pointing at everything,” Dr. Goldstein writes in the journal, “genetics would point at nothing.”
Two other geneticists, Peter Kraft and David J. Hunter of the Harvard School of Public Health, also writing in the journal, largely agree with Dr. Goldstein in concluding that probably many genetic variants, rather than few, “are responsible for the majority of the inherited risk of each common disease.”
But they disagree with his belief that there will be diminishing returns from more genomewide association studies.
“There will be more common variants to find,” Dr. Hunter said. “It would be unfortunate if we gave up now.”
Dr. Goldstein, however, said it was “beyond the grasp of the genomewide association studies” to find rare variants with small effects, even by recruiting enormous numbers of patients. He said resources should be switched away from these highly expensive studies, which in his view have now done their job.
“If you ask what is the fastest way for us to make progress in genetics that is clinically helpful,” he said, “I am absolutely certain it is to marshal our resources to interrogate full genomes, not in fine-tuning our analyses of common variations.”
He advocates decoding the full DNA of carefully selected patients.
Dr. Kraft and Dr. Hunter say that a person’s genetic risk of common diseases can be estimated only roughly at present but that estimates will improve as more variants are found. But that means any risk estimate offered by personal genomics companies today is unstable, Dr. Kraft said, and subject to upward or downward revision in the future.
Further, people who obtain a genomic risk profile are likely to focus with horror on the disease for which they are told they are at highest risk. Yet this is almost certain to be an overestimate, Dr. Kraft said.
The reason is that the many risk estimates derived from a person’s genomic data will include some that are too high and some that are too low. So any estimate of high risk is likely to be too high. The phenomenon is called the “winner’s curse,” by analogy to auctions in which the true value of an item is probably the average of all bids; the winner by definition has bid higher than that, and so has overpaid.
Dr. Kari Stefansson, chief executive of deCODE Genetics, an Icelandic gene-hunting company that also offers a personal genome testing service, said deCODE alerted clients to pay attention to diseases for which testing shows their risk is three times as great as average, not to trivial increases in risk.
Dr. Stefansson said his company had discovered 60 percent of the disease variants known so far.
“We have beaten them in every aspect of the game,” he said of rival gene hunters at American and British universities.
The undiscovered share of genetic risk for common diseases, he said, probably lies not with rare variants, as suggested by Dr. Goldstein, but in unexpected biological mechanisms. DeCODE has found, for instance, that the same genetic variant carries risks that differ depending on whether it is inherited from the mother or the father.