| dc.description.abstract | Paralleling the expansive growth of large-scale biobanks that contain patient electronic medical records-tied genome wide genetic data, the number of genome-wide association studies (GWASs) and the volume of newly identified disease associated genetic factors have expectedly accelerated in size as well. While this has facilitated an ongoing discovery process of countless disease-associated genetic loci, there remains an enormous gap in understanding the biological consequences and mechanisms related to these genomic areas. To address this gap, many bioinformatic methods including colocalization with functional annotations have been developed to better interpret GWAS findings. In this manner, many methods help identify expression quantitative trait loci (eQTLs), areas of the genome shown to regulate gene expression. Advancing this endeavor even further, several efforts have trained imputed gene expression models that capture the cumulative cis-regulatory effects on gene expression. Because of the immediate interpretability of these models, they can be utilized in regression models for transcriptomic wide association studies (TWAS) to characterize specific gene associations with phenotypes and diseases of interest. The work presented here largely strives to leverage this feature of interpretability by integrating these imputed gene expression models with other modeling methods to help design more interpretable genomic studies. By integrating with Bayesian frameworks, linear mixed effects models, and additive models, both unbiased and hypothesis-driven methods are demonstrated and characterized in the context of both common and rare diseases. This versatile collection of tools hopes to provide diverse means of approaching clinical phenotype by allowing for interrogation of both the entire genome and specific functional partitions of genetic factors. In addition to these transcriptomic-based tools, a method of quantifying cumulative ancestry-driven genetic effects on phenotypes is also demonstrated to help directly in interpretation of existing GWAS studies. Taken together, these methodologies both create approaches to better interpret existing results as well as provide tools for designing genomic investigations that possess immediate interpretability. | |
