Sequence Kernel Association Test, gene-environment interaction test, and meta-analysis for family samples with repeated measurements or multiple traits
MetadataShow full item record
Genetic loci identified by single variant association tests account for only a small proportion of the heritability for most complex traits and diseases. Part of the unexplained heritability may be due to rare variants and their interactions with environmental factors. Different strategies have been taken to increase the power to detect genetic associations, such as increasing the sample size by including related participants and meta-analyzing multiple studies. Longitudinal data or repeated measurements are often available in prospective cohort studies. For complex diseases, multiple traits are usually collected to characterize affected individuals. Many of the existing statistical methods can only be applied to the scenarios when each participant has one measurement of a single trait. To take full advantage of the data and further improve power, multiple measurements per individual may be included in the analysis when available. In this dissertation we develop statistical methods for rare variant association testing and gene by environment interaction analysis, and discuss gene-based meta-analysis for studies with different designs. First, we propose the generalized Sequence Kernel Association Test (genSKAT) to deal with rare variants, familial correlation, and repeated measurements or multiple traits. This is an extension of the original SKAT and family-based SKAT that accounts for correlation between multiple measurements within each individual. In the second part of this dissertation, we discuss methods to test for the presence of gene-environment interaction effects in the genSKAT framework. Finally, we evaluate genSKAT meta-analysis methods to combine different types of studies: samples of unrelated individuals with one observation per person or with multiple observations per person, and family samples with one observation per person or with multiple observations per person. Combining all these projects together, we contribute methodologies to detect rare variant associations by taking advantage of additional information, improve the chance to detect novel rare variant associations, and help in understanding the role that genetic factors play in various diseases and traits.