Now that the human genome has been mapped, one of the biggest challenges facing human geneticists is identifying groups of genes that collectively conspire to make some individuals particularly susceptible to a number of common diseases, including breast cancer, cardiovascular disease and depression.
Disentangling genetic predisposition from environmental factors is a highly complex process. So far geneticists have been able to do so only for a limited number of cases like cystic fibrosis that are caused by mutations in a single gene. Where more than two genes are involved, however, traditional methods of analysis have floundered because it has proven impractical to acquire genetic information from the large number of subjects required.
Now, however, a group of researchers at the Vanderbilt University Medical Center and the Vanderbilt-Ingram Cancer Center report that they have developed an alternative statistical approach to this problem. The technique, called Multifactor Dimensionality Reduction, can identify multiple gene interactions using data from a reasonable number of patients. Writing in the July issue of the American Journal of Human Genetics, the researchers report that they have used this technique successfully to identify four DNA sequence variations in three genes that work together to heighten a woman’s risk of breast cancer.
"For some time we have known that a person’s susceptibility to a number of common, complex diseases is not determined by a single gene, but by a number of genes working together," says Jason H. Moore, assistant professor of molecular physiology and biophysics, who led the research effort. "But, to the best of our knowledge, this is the first time that such a multiple-gene interaction has been identified."
Co-author and Professor of Pathology Fritz Parl, who has been studying the relation between estrogen and breast cancer for a number of years, predicts that this new approach will be widely used to study multiple-gene risk factors. When such an analysis is expanded to take non-genetic risk factors into account, it should significantly improve a doctor’s ability to determine the risk that certain treatments, like hormone replacement therapy, represent for individual patients, he says.
Most genes harbor common DNA sequence variations called polymorphisms and rare DNA sequence variations called mutations. Some mutations single-handedly increase an individual’s susceptibility to specific diseases. Such "genetic" diseases are recognizable because they are heritable and so cluster in certain families. An example is hereditary breast cancer, which accounts for less than 10 percent of all breast cancer cases. It is highly associated with the action of mutations in one of two genes.
On the other hand, most common diseases do not exhibit a clear pattern of heritability. So geneticists argue that the observed variations in susceptibility must be caused by the interactions among multiple polymorphisms. In such cases, individual polymorphisms are harmless, but when they occur in a specific combination they significantly enhance a person’s risk. In some cases, the increased susceptibility may be due to the collective action of a few polymorphisms, but in others they may arise from the subtle interactions among hundreds of gene-variants.
Take the case of the polymorphisms that the Vanderbilt group has linked with the sporadic breast cancers that occur in women with no family history of the disease and account for more than 90 percent of all breast cancer cases.
The researchers began by looking at five genes involved in estrogen metabolism. They chose this particular set of genes because there is considerable evidence that estrogens influence breast cancer risk and recent studies have shown that the enzymes that break down estrogen in the body produce metabolites that can cause cancer.
Moore, graduate student Marylynn Ritchie and programmer Lance Hahn analyzed 10 functional polymorphisms that alter the levels of the suspect estrogen metabolites. When they looked at the polymorphisms individually, he and his colleagues found no indication of increased cancer risk. It was only when they looked at different combinations that they found that women with four specific polymorphisms were significantly more likely to develop breast cancer than those with only three, two or one of these gene variants.
They found this complex association by applying their technique to the genetic information that Parl had compiled on 200 women with breast cancer and an age-matched group of 200 female patients without the illness. First, they constructed a series of tables that compare pairs of polymorphisms. Because a person can inherit a given polymorphism from her father, mother or both parents, each table has nine cells. In each cell, they calculate the number of subjects with breast cancer and the number of control subjects without the illness who have the indicated pair of polymorphisms. If the number with cancer is higher, then the cell is listed as high risk; if the number without cancer is higher, then it is considered low risk. If none of the subjects have a particular combination of the two polymorphisms, the cell is labeled as empty.
The researchers constructed these tables for all possible pairs of polymorphisms. Then they constructed similar tables for all combinations of three, four, five … all the way up to nine polymorphisms. By examining the pattern of high and low risks in these tables, the geneticists determined that only one combination—four polymorphisms in three genes—increased an individual’s cancer risk. Moore’s approach isn’t limited to analyzing interactions among three or four genes. Using machine learning algorithms and a new multi-processor supercomputer on campus, he estimates that they will be able to search for similar interactions among as many as 20 genes selected from a list of thousands of candidates.
Other members of the research team are technician Nady Roodi and analyst L. Renee Bailey who now works for AstraZeneca in Wilmington, DE.
The research was funded by the National Institutes of Health, the Vanderbilt-Ingram Cancer Center and the Vanderbilt University Medical Center. With the help of Vanderbilt’s Office of Technology Transfer Moore has decided to make a Multifactor-Dimensionality Reduction software package freely available to academic users. For more information is available online at http://phg.mc.vanderbilt.edu/Software/MDR.
The paper is available online at http://www.journals.uchicago.edu/AJHG/journal/contents/v69n1.html.
American Journal of Human Genetics