Title
Two-locus modeling of complex traits with large complex pedigrees and marker data
Speaker
Yuqun Luo, The Ohio State University
Abstract
Modeling the underlying genetic mechanism is an important step in mapping genes underlying complex genetic traits. Many common and chronic human diseases, such as some forms of cancer, diabetes and asthma, are complex genetic traits, involving multiple genes, environmental risk factors, and possibly their interactions. It is vital in modeling complex genetic traits to incorporate data from genetic markers (genetic material of known chromosomal positions) to improve accuracy in estimation of model parameters. Furthermore, as a first step to tackle the complexity, it is natural to consider two-locus disease models. It is also advantageous to use a reproductively isolated founder population, where relative homogeneities in genetic materials and environmental exposure eliminate problems caused by admixture populations. However, pedigrees from such a population are often very large and complex, with inbreeding among related individuals. The common practice of breaking such pedigrees into smaller, simpler pieces in modeling will lose much information.
The first part of the talk is focused on the information gains from incorporation of marker data in the estimation of disease model parameters, to warrant the extra effort exerted on such incorporation. We show that substantial variance reductions are achieved as a result of the incorporation, and the reductions are greater with more polymorphic marker(s) and larger pedigrees. In the second part of the talk, we present an application of a Bayesian Markov chain Monte Carlo methodology for analyzing an asthma dataset from the Hutterite population. The 1544-member pedigree is analyzed as a whole with incorporation of marker data. Among the competing classes of one-locus and two-locus models, an epistasis dominant-dominant model fits the data best in terms of population prevalence and relative risks.
