Novel Approaches in Linkage Analysis for Complex Traits

To develop new statistical methods to explore genetic mechanisms that contribute to the development of hypertension.

BACKGROUND: Hypertension affects 50 million Americans and is the single greatest risk factor contributing to diseases of the brain, heart, and kidneys. There is a strong evidence that hypertension has a genetic basis. The study will develop novel approaches to better understand the genetic mechanisms contributing to measures of blood pressure (BP) level, diagnostic category (hypertension versus normotension) and correlated traits. DESIGN NARRATIVE: This genetic epidemiology study will develop novel approaches to better understand the genetic mechanisms contributing to measures of blood pressure (BP) level, diagnostic category (hypertension versus normotension) and correlated traits. The first aim is to localize genes influencing measures of blood pressure levels, diagnostic category and their correlates. This will be done by applying genome-wide multivariate linkage analyses based on the variance components approach and utilizing clusters of traits correlated with measures of blood pressure and/or diagnostics category. The second aim is to develop exploratory diagnostic tools for linkage analysis of complex traits to further enhance our ability to localize genes influencing measures of blood pressure, diagnostic category and their correlates. This will be done by extending the diagnostic tools used in regression analysis to the variance components approach used for linkage analysis of quantitative traits. In this study for example, it can be used to identify outlier families since previous studies have shown that families with outlier values yield false-positive results. Tree-structure models will also be extended to pedigree data. Tree-based modeling is an exploratory technique for uncovering structure in the data. The use of tree-structure models is advantageous because no assumptions are necessary to explore the data structure or to derive parsimonious model. These models are accurate classifiers (binary outcome) and predictors (quantitative outcomes). All these tools will be incorporated in the S-Plus software as a function. S-Plus was selected due to its capability and flexibility for analyzing large data sets.

Olswold C, de Andrade M. Localization of genes involved in the metabolic syndrome using multivariate linkage analysis. BMC Genet. 2003 Dec 31;4 Suppl 1(Suppl 1):S57. doi: 10.1186/1471-2156-4-S1-S57.

Fridley B, Rabe K, de Andrade M. Imputation methods for missing data for polygenic models. BMC Genet. 2003 Dec 31;4 Suppl 1(Suppl 1):S42. doi: 10.1186/1471-2156-4-S1-S42.

Pankratz VS, de Andrade M, Therneau TM. Random-effects Cox proportional hazards model: general variance components methods for time-to-event data. Genet Epidemiol. 2005 Feb;28(2):97-109. doi: 10.1002/gepi.20043.