Genetic association studies in the presence of different family types: a meta-analytic approach

Family-based designs represent an important class of studies used in genetic association to locate genes underling complex diseases. A reference family type for these designs is a family trio, where an affected individual and his/her parents are genotyped at markers of interest. However, the collection of a large number of trios can be unfeasible for several reasons, such as, for example, small population size, unavailability of parents or missing genotype data. Thus, available individuals from the proband's pedigrees other than parents, typically siblings or spouses, are occasionally genotyped, leading to samples with a mixed family structure. This thesis investigates and suggests methods to perform association analysis on samples containing different family types. Two meta-analytic strategies for analysis are suggested to conduct single-locus association tests. First, a genetic application of the well-known stratified Cochran-Mantel-Haenszel test, which allows for the analysis of trios, sib-pairs and unrelated affected and unaffected individuals, is extended to diads, i.e., one affected child-one parent pairs. Second, different association tests from the literature are conducted on the stratum of trios, the stratum of diads, the stratum of discordant sibships and the stratum of unrelated affected and unaffected individuals respectively, and then the resulting p-values are combined with meta-analytic techniques. The two strategies are shown to work effectively on a real dataset from a family-based association study on Multiple Sclerosis in Central East Sardinia. Comparisons with results from current software are presented, together with guidelines for usage. The methods proposed in this thesis mainly work on the complete data only. Additional work on the relative power of the two strategies is required, along with comparisons with other methods which accommodate the unavailability of individuals and/or missing genotype data