Novel methodologies for genetic association testing
This work presents new contributions to genetic association testing, for both family-based and population-based designs. In the first chapter, the current genome-wide screening methodology for quantitative traits in family-based designs is described and applied to a study on alcoholism. This methodology allows for both screening and replication within the same data set. Using this technique, an association is detected between a marker on chromosome 4 and alcoholism-associated electrophysiological phenotypes.
While the current screening methodology outlined in Chapter 1 has been successful in identifying potential genetic determinants of complex disease (Van Steen et al., 2006; Herbert et al., 2006), it is restricted to the analysis of quantitative traits. In the second Chapter, a novel two-stage testing strategy is presented that may be applied to study designs where only affected probands are analyzed. The power of the approach is estimated via simulation studies, and its practical relevance is illustrated by an application to a partial genome-scan of candidate genes for asthma that identifies three SNPs in a 6KB region.
One critique of this new two-stage screening and testing methodology is that the screening step is potentially susceptible to population admixture. For the screening of quantitative traits, genomic-control techniques (Devlin and Roeder, 1999) may be employed. However, due to the construction of the test statistic for screening affected probands, genomic control techniques may not be directly applied. In researching test statistics that may accommodate genomic-control measures, a new test statistic for measuring Hardy-Weinberg disequilibrium was discovered and developed. In the third Chapter, the new test is contrasted against the standard methodology for testing for Hardy-Weinberg disequilibrium, and it is shown that it has improved power in most testing scenarios.
Finally, in population-based association testing, new methodology is developed for sample-size and power calculations for matched case-control designs. Within the framework developed by Gauderman (2002), the current methodology is extended to allow for environmental correlation between case and control environmental exposures, ordinal environmental exposure variable, and multiple controls for each case. The results demonstrate that environmental correlation has a minimal impact on sample size requirements for detecting gene-environment interactions, decreasing requirements for most scenarios.