Description
- Abstract:
- Statistical genetics has improved our understanding of how genetic variation shapes human complex traits and diseases. Growing biobanks, improved sequencing technologies, and efforts of scientific collaboration have enabled the field to make progress in estimating heritability, detecting single-variant associations, and elucidating polygenic architectures. Genome-wide association studies (GWAS), rare variant tests, and models for gene-by-gene (GxG) and gene-by-environment (GxE) interactions have established a foundation for translating the insights from statistical genetics to clinical risk prediction and drug target discovery. Despite ever-increasing scale and availability of data, many questions about the genetic architecture of complex traits remain. Traits may be driven by thousands of variants with small effects, while linkage disequilibrium patterns, pleiotropy, epistasis, and heterogeneity in environmental exposures obscure genetic inference. GWAS signals frequently are located in non-coding regions, complicating biological interpretation. Many methodological challenges remain in order to better understand nonlinear contributions to phenotypic variation, disease susceptibility, and frailty and to bridge the gap between statistical associations and biological causality. The present dissertation addresses these challenges by extending the marginal interaction framework to multivariate analysis and time-to-event traits, and by providing a framework for integration of multi-omic information into genetic association studies. Chapter 1 introduces a multivariate marginal epistasis test (mvMAPIT) that leverages genetic correlations between traits to detect genetic variants involved in GxG interactions. This method is implemented as a multivariate linear mixed model and improves power for detecting marginal epistasis in multivariate data. Chapter 2 presents the Cox proportional hazards gene-by-environment interaction test (CphGxE). This model enables the detection of GxE interactions in time-to-event traits by partitioning the heritable variance of frailty into genetic, environmental, and GxE interaction components. Chapter 3 proposes the Sparse Marginal Epistasis (SME) test. This approach enables the integration of functional data as biological priors into genome-wide association studies of epistasis, achieving substantial improvements in power and scalability of the marginal interaction framework to biobank-scale data. Together, these methods present computational and statistical advances by providing scalable and interpretable approaches for studying nonlinear genetic architectures in complex traits.
- Notes:
- Thesis (Ph. D.)--Brown University, 2025
Citation
Stamp, Julian,
"Statistical and Computational Advances for Detecting Nonlinear Contributions to Complex Traits"
(2025).
Center for Computational Molecular Biology Theses and Dissertations.
Brown Digital Repository. Brown University Library.
https://repository.library.brown.edu/studio/item/bdr:zavpcrh5/
Relations
Collection:
-
Center for Computational Molecular Biology Theses and Dissertations
Theses and Dissertations for the Center for Computational Molecular Biology department....