Search

A substantial amount of effort has been expended recently towards the identification and evaluation of tag single nucleotide polymorphisms; markers that, due to linkage disequilibrium (LD) patterns in the genome, are able to act as "proxies" for other polymorphic sites. As such, these tag markers are assumed to capture, on their own, a large proportion of the genetic variation contributed by a much greater number of polymorphic sites. One important consequence of this is the potential ability to reduce the cost of genotyping in an association study without a corresponding loss of power. This application carries an implicit assumption that strong LD between markers implies high correlation between the accompanying association test results, so that once a tag marker is evaluated for association, its outcome will be representative of all the other markers for which it serves as proxy. We examined this assumption directly. We find that in the null hypothesis situation, where there is no association between the markers and the phenotype, the relationship between LD and the correlation between association test outcomes is clear, though it is not always ideal. In the alternative case, when genetic association does exist in the region, the relationship becomes much more complex. Here, reasonably high LD between markers does not necessarily imply that the association test result of one marker is a direct substitute for that of the other. In these cases, eliminating one of these markers from the set to be genotyped in an association study will lead to a reduction in overall power., ((c) 2008 Wiley-Liss, Inc.)

The simulated sequence data for the Genetic Analysis Workshop 12 were analyzed using data mining techniques provided by SAS ENTERPRISE MINER Release 4.0 in addition to traditional statistical tests for linkage and association of genetic markers with disease status. We examined two ways of combining these approaches to make use of the covariate data along with the genotypic data. The result of incorporating data mining techniques with more classical methods is an improvement in the analysis, both by correctly classifying the affection status of more individuals and by locating more single nucleotide polymorphisms related to the disease, relative to analyses that use classical methods alone.