|Institution:||University of Washington|
|Full text PDF:||http://hdl.handle.net/1773/9532|
Linkage disequilibrium (LD) is the dependence of alleles at different loci in the population. Understanding the patterns of linkage disequilibrium in humans has been of great interest lately because of its potential impact on the mapping of disease genes and the study of population history. In this dissertation we propose a new statistical model for linkage disequilibrium that relates patterns of LD to the underlying recombination rate. The model is based on approximating the conditional distribution of observing a new haplotypes given existing haplotypes and hence approximates the full data coalescent likelihood directly. One major advantage of this new model is its computational efficiency. We first applied this model to estimate constant recombination rate with simulated data. With an empirical correction in order to correct the bias, our method performed very well compared with existing methods. We then used two models, the single-hotspot model and the general variable rate model, to detect the recombination hot spots and explore the fine-scale heterogeneity of recombination rate. Simulation studies confirmed the power of this approach in detecting recombination hot spots. Using real data, our analysis confirmed previous findings based on sperm-typing and other analytic models. Because our method uses full haplotype data, we were also able to discover some new interesting pattern. Finally we extended the model to case-control data and fine-scale association mapping. Preliminary results were encouraging but there is still issues that need to be addressed in the future to improve the efficacy of the method.