Collections > Electronic Theses and Dissertations > Making Robust Use of Parental Genotype Data for Finding Effects of Variants on the X Chromosome
pdf

The X chromosome is generally understudied in association studies, in part because the analyst has limited methodological options. We are developing statistical methods for causal association for single nucleotide polymorphisms (SNP markers) on the X. The focus of our work is on case-parent triad association studies. Most current family-based methods extend the transmission disequilibrium test (TDT) to the X chromosome. We propose a new method to study association in case-parent triads: the parent-informed likelihood ratio test for the X chromosome (PIX-LRT). Our method provides estimation of relative risks and takes advantage of parental genotype information and the sex of the affected offspring. Under a parental allelic exchangeability assumption for the X, if for a given locus case-parent triads are complete, the parents of affected offspring provide an independent replication sample for the estimates based on transmission distortion to the affected offspring. For each offspring sex we can combine the parent-level and the offspring-level information to form a likelihood ratio test statistic; we then combine the two to form a single composite test statistic, which we show offers better power than existing methods. Maternal SNP effects can influence the development and later health of the offspring through prenatal effects, regardless of which alleles are transmitted by the mother to her offspring. Previously, using triads alone, no method had been developed without an assumption of Hardy-Weinberg Equilibrium (HWE) to test maternal effects on the X chromosome. For the second project we extended PIX-LRT to discover maternal X-chromosome SNP effects. Our third project concerns the identification and estimation of effects of X haplotypes. For case-parent triads, the X-chromosome haplotype phases can be inferred. With phase information, as is available when triad genotypes are nonmissing, the problem can be managed via an extension of the PIX-LRT from a two-allele problem to a k-allele problem, where the “alleles” are now the existing haplotypes at the locus under study. The extended approach relies on a permutation-based p-value based on the most significant individual haplotype effect. Our methods are applied to a dataset consisting of over 2000 triads in which the affected offspring have an oral cleft.