Robert Elston

Credited with helping found the field of statistical genomics, Elston received this year’s Marvin Zelen Leadership Award. Zelen is a professor of statistical science at HSPH who has created probability models of disease; has developed new statistical methods for the planning and analysis of clinical trials; and has suggested a theory for the early detection of disease. Recently, Zelen has found that the recent decrease in breast cancer deaths is mostly due to early detection and widespread use of supplemental treatment, especially tamoxifen.

Elston, a professor in the Department of Epidemiology and Biostatistics at Case Western Reserve University, previewed a new way of finding disease susceptibility genes in large data sets of unrelated people. "This is the way that the field of genetics is moving, away from family to case-control," he said in a follow-up interview.

For example, on one small microarray chip, cancer researchers now can analyze thousands of genes from one person. By comparing hundreds of such chips from unrelated people, researchers may be able to find common patterns that help define new subsets of disease, predict which people will respond to medications, and identify the origins and processes of disease.

One of the things statisticians have to worry about in such analyses is "false positives," or finding apparent connections between genes and disease where none exists. Unfortunately, that can easily happen. Mathematically precise genetic theory is often based on the assumption of many generations of random mating to pass along genes. In real life, many groups, such as Irish Catholics or Ashkenazi Jews, have tended to marry among themselves, Elston said. In that case, tantalizing clusters of genes might merely be markers of certain communities, not disease processes.

To cope with this challenge, Elston and Kijoung Song, a postdoctoral fellow in Elston’s lab, propose combining two methods of searching for candidate genes involved in complex diseases. One method looks for the frequency of a gene variation–or an allele–in people who have a disease, compared to people in a similar population who do not. The other method compares the proportions of a diseased population with no copies or multiple copies of a certain gene to the theoretical frequencies expected after many generations of random mating. Elston and Song showed that this combination method worked in avoiding false positives for a simulation of 500,000 people, but added that more research is needed for a strict proof.

--CCM