Tools for cluster analysis of data from genome-wide association studies

In the past couple of decades, genome-wide association studies (GWAS) have become a widely used approach for investigating the underlying genetic architecture of complex human diseases. Each particular GWAS will highlight multiple loci across the genome, in which genotypes occur at significantly different frequencies in cases versus controls. These loci are referred to as risk variants, and are believed to have some effect on the disease under investigation. A primary goal of GWAS is to gain insight into the underlying etiology of the diseases subject to study. In this setting, one area of interest would be understanding how and to what extent diseases and traits are related on the genomic level. Cluster analysis, a machine learning technique that aims to group similar objects within a data set, can be used for the purpose of explorative analysis of the relationships between disease-associated data. However, the results and interpretation of a cluster analysis relies considerably on choices of data representation and distance measure, the details of which have not been thoroughly discussed in the context of GWAS data sets. This thesis presents different methods for cluster analysis of disease- associated genomic tracks. Different ways to uniformly and computationally represent GWAS data are discussed, where the biological property linkage disequilibrium is taken into account. Further, we describe objective measures for defining similarity with respect to the different data representations. The goal of these different methods is to reflect a biological, rather than technical notion of similarity between disease-associated data. A suite of tools and methods for cluster analysis of GWAS data sets is made available through the Genomic HyperBrowser, an open-source web-based analysis platform developed and maintained by the research group for biomedical informatics at the University of Oslo. The tools aim to provide a comprehensive and transparent way to explore the extent to which diseases share genetic components identified from GWAS. Examples of utilization, with published workflows and data sets, are provided