| 總結: | <p>Epistasis (gene-gene interaction) is a universal component of common complex genetic traits. The identification and characterization of epistatic interactions are crucial to a full understanding of the complex genetic mechanisms that underlie human disease. The aim of this thesis is to examine epistasis in non-parametric linkage analysis of human complex traits, with an emphasis on the affected sibling pair (ASP) study design.</p><p>Following an overview of approaches that model and detect epistasis in linkage analysis of human complex traits, I present an extension of a two-locus nonparametric linkage method in ASPs. The new two-locus approach, Merloc, jointly models pair-wise interactions between susceptibility loci in different types of affected relative pairs and estimates of the most likely underlying genetic model for a pairwise interaction, implemented to genome-wide applications. To test the performance of the approach, Merloc was compared to two multilocus non-parametric conditional linkage approaches. Power and type I error rates under null, single-locus, and twolocus genetic models of epistasis and heterogeneity indicated that Merloc outperformed the other methods.</p><p>The method was applied to type 2 diabetes data to assess the evidence for epistasis between two susceptibility loci. Significant evidence for epistasis was obtained supporting previous findings from conditional interaction analysis. A search through the space of parametric two-locus models indicated that nine two-locus models best approximated the pair-wise interaction.</p><p>Genome-wide strategies to detect epistasis were also examined in this thesis and the simultaneous search for genome-wide interactions was explored in detail. Two-dimensional (2D) linkage scans were performed using Merloc in three complex traits, essential hypertension, autism, and type 2 diabetes. Several peaks were detected in the two-dimensional likelihood surfaces with genome-wide suggestive evidence for linkage. Extensive simulations were used to examine the distribution of the test statistic under the null hypothesis in the context of two-dimensional linkage scans.</p><p>Finally, two main extensions of this approach were considered - linkage approaches to examine more than two loci, and extending the method in this study to include a test of association.</p>
|
|---|