Department of Statistics Seminar
North Carolina State University

presents

Ina Hoeschele

Virginia Polytechnic Institute and State University

"Bayesian Gene Mapping for Complex Pedigrees and Genetic Parameter Estimation in Finite Polygenic Mixed Models"

ABSTRACT

A Bayesian method for mapping genes affecting quantitative traits (QTLs) in complex pedigrees is presented. The Bayesian method was chosen because it accounts for all the uncertainty in the system, including unknown QTL and marker genotypes as well as number of QTLs. Furthermore, it allows the investigator to incorporate biologically meaningful prior information. As the number of genes to be mapped simultaneously is unknown, Bayesian inference involves model choice and model averaging, with the model representing the number of QTLs. Bayesian analysis is implemented via Markov Chain Monte Carlo (MCMC) algorithms. Two techniques are compared which allow the Markov chain to jump between model subspaces, a reversible jump sampler and a technique related to the product-space sampler with pseudo-priors. Quantitative genetic models partition the phenotype into genetic and environmental components, with the former further partitioned into additive, dominance, additive-by-additive, etc., components. In a finite polygenic model (FPM), the genetic component is modeled discretely as the sum of the genotypic deviations at a finite number of individual loci. The FPM is being investigated as an alternative to the infinitesimal model with particular consideration given to the estimation of additive, dominance, and additive-by additive genetic variance components.

Monday, October 26, 1998

8:00 - 9:00 am

2405 Williams Hall

Coffee and donuts will be served at 7:45 am.