Statistical Methods for Detecting Rare Variant Associations in Family-Based Design

Rare variants are hypothesized to explain some genetic contribution to complex traits. However, using conventional case-control designs to identify rare variants associated with traits has low statistical power. Family designs can substantially increase power for these studies, especially for rare variants. In this dissertation, we present innovative statistical methods that can efficiently use family information to improve power.;In Chapter 2, we present TRAFIC, a rare-variant association test using affected sibpairs. For rare risk variants, two affected siblings would share the variant on their shared identity-by-descent (IBD) chromosomes. We thus test the distribution of rare variants on IBD chromosomes and non-IBD chromosomes. TRAFIC is robust to population stratification as "cases" and "controls" are matched within each sibpair. We show that TRAFIC has significant power gain over the population case-control design for variants with summed allele frequency < 5%. Considering allelic heterogeneity, where risk variants have different effect sizes, TRAFIC can double the power of a case-control study.;In Chapter 3, we present TRAP for testing the association between rare variants and a binary trait using extended families. Since affected family members are more likely to share risk variants, we propose to test if rare variants are shared more than expected given the known inheritance vector and the founder genotypes. TRAP is applicable from small to large pedigrees with multiple generations, including families with missing founders. Using simulations, we show that TRAP is more powerful than the conventional case-control design and existing family-based approaches, especially for rare variants.;In Chapter 4, we present testing for rare variants associated with a continuous trait in extended families (TRACE). Given that a rare variant increases the trait value, conditional on inheritance vectors and founder genotypes, we test if family members with high trait values are more likely to share the variant. Under non-ascertained scenarios, TRACE can be more powerful than the existing family-based methods for large pedigrees; for ascertained scenarios, TRACE outperforms the existing approaches throughout considered pedigree structures. In sum, we present three family-based methods efficiently using the sharing of variants to increase the power for detecting rare variant associations.