Autism Research Made to Order
Freeman, Kris, Environmental Health Perspectives
Over the past decade, many studies have suggested that the genetic risk for autism is related to several genes, but identification of a known autism susceptibility gene has eluded scientists. Now, using a new statistical method known as ordered-subset analysis (OSA), researchers at the Duke University Center for Human Genetics have linked one type of autistic behavior to a specific gene (GABRB3) on chromosome 15. With this study, the researchers have both narrowed a region of interest for future autism studies and shown that OSA is an effective means for mapping disease-susceptibility genes.
"The use of OSA in autism represents just one effort to adapt an analytical strategy in a new and exciting way to maximize the information we can extract from our data," says Margaret Pericak-Vance, center director and principal investigator of the study, which was published in the March 2003 issue of the American Journal of Human Genetics.
Autism, estimated to affect some 1.5 million Americans, has been associated with a number of genes, giving rise to a multitude of variables and making the study of genetic links to autism exceedingly complex. "If only one gene contributes to a disease, all the families you study will have variations of that gene," explains Ellen Wijsman, a research professor of medicine and biostatistics at the University of Washington. "But if two or three or ten genes are involved, a very small number of families may have variations in a given gene. You may need very large sample sizes to pick up these very weak genetic signals."
Another problem arises in deciding how to group participants for statistical analysis. "When one is trying to subdivide a sample of families on the basis of a continuous covariate such as [child's] age or severity of behavior, it can be very difficult to choose and defend cut points," says Elizabeth Hauser, a statistical geneticist at the Duke center who helped develop the OSA method. Taking an example where age of onset might be analyzed as a variable, she asks, "Is the genetic effect likely to be strongest in the half of families with the youngest age of onset, or the third of families with the youngest age, or the quarter of families with the youngest age?"
Instead of comparing predefined groups, OSA ranks participants on a continuum and automatically selects the group that provides the best match, for example of a trait to a particular gene. The OSA method was built on a technique used in the mapping of the breast cancer gene BRCA1. …