Omics Data Exploration: Across Scales and Dimensions.

Abstract

The rapid development and adoption of high throughput technologies has led to an avalanche of omics data, including those from genome, transcriptome, proteome and metabolome, from individual laboratories as well as global-scale collaborative efforts. The major ensuing challenge is then how to analyze, explore and extract new biomedical knowledge from such omics datasets. This thesis attempted to address some of these challenges by 1) developing novel tools for flexible searching, clustering and visualizing omics networks and pathways 2) developing novel robust statistical workflows to identify confident associations that lead to discovery of new cell-line specific bio-signatures from NCI-60 omics datasets with high variability and missing measurements, and most notably, 3) conceiving and developing a novel visual data exploration model, the CoolMap, to bring multi-scale, versatile and flexible visual data mining capabilities to structured two-dimensional omics datasets. CoolMap’s unique capabilities were demonstrated through several use cases including a mother-child nutrient/epigenetics study, and enables efficient and flexible identification of strongly correlated high-level ontological concepts as well as low-level specific measurements for data-driven hypothesis generation.