A basic tenet of statistical genetics is that shared ancestry underlies population associations between complex traits and genetic markers such as haplotypes. The coalescent is a popular mathematical model of the ancestral tree of copies of a genetic locus sampled from the population. While these ancestral trees cannot be observed directly, the observed haplotype data enable sampling from their posterior distribution. For each sampled tree, statistical tests of shared ancestry among extreme phenotypes may be applied to summarize the trait association. The distribution of p-values from these association tests is a fuzzy p-value, as defined by Thompson and Geyer. This talk will review some basic ideas behind association fine-mapping of complex trait loci and discuss the utility of coalescent ancestries and fuzzy p-values in this regard. I will present some preliminary results from ongoing experiments with these ideas on an example dataset of 125 cases and 125 controls defined by an immune marker predictive for type 1 diabetes. This is joint work with Kelly Burkett and Brad McNeney.
Jinko Graham is an Associate Professor of Statistics and Actuarial Science and a member of the Statistical Genetics Working Group at Simon Fraser University. Her research is directed towards the development of statistical methods for inference from genetic data, with a focus on genetic association studies. One question of interest is incorporating the gene genealogies underlying genotypes at multiple markers to map the genomic location of traits under genetic influence. Another question is the unbiased display and inference of statistical interaction between genes and the environment in complex diseases of early onset, using data from cases and their parents. Much of her research involves complex data structures and so has a strong computational component. Dr. Graham is enjoying time in Montreal, working with Dr. Celia Greenwood and colleagues.