Statistical methods for analyzing and interpreting high-dimensional phenotypes

Dimension reduction techniques, such as singular value decomposition and independent component analysis among others, have been instrumental in delineating complicated correlation structure and separating signal from noise in multiple systems. Biological data in the genomic age encompasses everything from the genome to the transcriptome, the metabolome, the methylome, the proteome and more. Such experimental platforms typically generate millions to billions of data points making them ideal for the methods mentioned above. In this dissertation, I will outline my work on utilizing dimension reduction methods in elucidating novel biological insights by applying them to genome-wide polymorphism and transcriptome profiling data, thus defining correlated modules of genetic variants and gene expression respectively. In the first study, I use segregants derived from crosses generated between two divergent Saccharomyces cerevisiae strains to construct meta-phenotypes and then perform genome-wide linkage analyses to identify novel genomic loci modulating phenotypic variation. In the second study, I apply multivariate dimension reduction techniques in the context of single nucleotide polymorphisms genotyped in unrelated individuals sampled from diverse human populations, to define fine-scale patterns of population structure within continents both at global and local scale along the genome. The genomic distribution of correlated variants facilitates inferences regarding the stochastic effects of genetic drift while evidence for adaptive evolution is supported by findings of significant clustering. In the final study, I analyze the transcriptional response induced due to innate immune system activation in B-lymphoblastoid cell lines isolated from European American and Yoruban African parent-offspring trios. Administering a specific immune challenge leads to widespread differential expression of genes and a significant fraction of the change in expression is heritable. In addition, I identify genetic loci that are associated with the heritable variation in transcriptional response, which provide us with insights into the mechanisms of the genetic control of gene expression in perturbed systems, specifically in the innate immune related pathways.