ABSTRACT; Period 2007-2011

Main | Personnel | Software | Reports

Techniques will be developed for genetic analysis of complex diseases segregating in extended pedigrees of arbitrary structure. Cardiovascular, neurological and behavioral traits are among those having both environmental and genetic components. However, identification of genes contributing to increased risk of related disorders has been limited by both computational and statistical constraints. The development of Markov chain Monte Carlo (MCMC) methods has helped to overcome these limitations, providing information for gene localization, trait model estimation, haplotype analysis, and genetic map analyses, using data on extended pedigrees. The research now proposed is concerned with the extension of MCMC methods in several areas. Improved methods will be developed for the MCMC analysis of gene descent in extended pedigrees, given data at a dense genome screen of markers, together with the use of these gene descent patterns in joint multilocus linkage and segregation analysis of complex traits. Models for discrete and quantitative trait phenotypes in these analyses will be extended to include epistasis and pleiotropy. MCMC-based likelihood-ratio methods for assessment of trait-model robustness will be incorporated into our toolkit. Further, methods will be developed for assessment of the statistical significance of linkage findings, including correction for multiple testing at linked genome locations. Conditional on marker data, trait-data resimulation and permutation will be used to develop measures of significance. Also, the conditional distributions of gene descent at marker locations will be used to provide both measures of significance and confidence sets for trait-locus locations. Methods will also be developed for the use of patterns of gene descent realized conditional on marker data in the analysis of genetic maps and marker models, including multi-SNP haplotypes and copy-number variants in these analyses. The impact of marker map and model uncertainty on linkage findings will be investigated. In using multiple dense SNPs as markers, methods for incorporating linkage disequilibrium (LD) into the analysis of family data will be developed. The impact of LD on MCMC-based methods of haplotype inference, estimation of identity by descent ({\ibd}) over regions, and lod score analyses will be investigated. Methods will be evaluated by analyses on several simulated and real data sets, including pedigrees segregating cardiovascular disease or behavioral disorders. These real data sets include several on which are available genome-wide marker screens or more localized multigene haplotypes. Finally, software will be developed that implements these methods, and will be documented and released for use by practitioners.
UW - Statistics: Thursday, 13-Dec-07 Contact: Elizabeth Thompson <eathomp@u.washington.edu>