Genetics

Edited by: Ricki Lewis E. Lander, L. Kruglyak, "Genetic dissection of complex traits: Guidelines for interpreting and reporting linkage results," Nature Genetics, 11:241-7, 1995. (Cited in more than 182 publications through August 1997) Comments by Eric Lander, Whitehead/MIT Center for Genome Research MAKE 'EM LAUGH: Eric Lander used humor to counterbalance heavy use of statistics in his paper on establishing linkage between a complex trait and genetic markers. For decades, geneticists have attempted to link the genes controlling human traits to chromosomes. At first, linkage data associated inheritance of an abnormal chromosome with a blood type or other detectable trait. In the 1980s, the first restriction fragment-length polymorphism (RFLP) maps correlated phenotypes, such as an inherited disease, to alternative DNA sequences (polymorphisms) that changed cutting sites for restriction enzymes. Today, the human genome is riddled with so many genetic markers that investigators have ventured beyond the predictably inherited Mendelian traits to the study of complex traits. These characteristics "run in families," but not in the frequencies that Mendel's laws predict. Complex traits are typically caused by more than one gene or by environmental influences. This paper by Eric Lander, a professor of biology at the Massachusetts Institute of Technology and director of the Whitehead/MIT Center for Genome Research, and Leonid Kruglyak, a research scientist at the Whitehead Institute for Biomedical Research in Cambridge, Mass., provides guidelines to evaluate the strength of linkage between a complex trait and genetic markers. Establishing linkage entails determining how often blood relatives share a phenotype of interest and a particular DNA marker. Today, scientists search for such correlations across the whole genome. Statistics used to demonstrate linkage of a trait to just one marker or chromosomal region must be robust enough to support whole genome screens. In other words, if a researcher is looking for coinheritance of a trait with a particular short DNA sequence, the sequence is more likely to be found if the whole genome is searched than if only one chromosome segment is considered. "There's a difference between a point-wise test, which is a single experiment, and a genome-wide search, in which you scan the entire genome," Lander explains. "A genome-wide scan is the equivalent of hundreds of experiments. This paper asks, 'If you look everywhere in the genome, what is the highest correlation that you will see just by chance?' That concept hadn't been firmly embedded in the literature." Geneticists use two easily confused measurements to describe linkage-a logarithm of the odds (LOD) score and a P value. "The basic problem is that a LOD score indicates how much more likely a hypothesis of linkage is to be true than the alternative hypothesis of nonlinkage," Lander says. "By contrast, a P value tells how often one would see the observed evidence by chance. A LOD score of 3-the standard for determining linkage for a Mendelian trait-means that the data are 1,000 times more likely to arise from linkage than from nonlinkage, but this doesn't mean that the odds of linkage are 1,000 to 1. It's a subtle but crucial distinction." Lander and Kruglyak suggest using LOD scores and P values for complex traits that are slightly more stringent than those accepted as evidence for linkage for Mendelian traits. They propose four levels of stringency for estimating the robustness of linkage for complex traits-suggestive, significant, highly significant, and confirmed. The goal was to provide a clear way for the reader of a paper to "pigeonhole the result-is the evidence overwhelming, or merely suggestive? That is, could the evidence have arisen by chance?," Lander says. He is surprised that a paper on linkage statistics would be highly cited but acknowledges that the work is important to those tracking down the genetic origins of complex traits. "Dozens and dozens of papers published on these characteristics cite this paper because it provides guidance for the analyses. It is an intellectual infrastructure of sorts," he comments. The article is also intriguing for its humor. The authors, for example, write that "biologists often greet statistical issues with glazed-eyed indifference." Concludes Lander: "It is a very critical question to figure out how common complex traits are inherited, and that involves statistics. The humor in the article helps the medicine go down."

Genetics

Interested in reading more?

Become a Member of