Methods and tools in computational human genetics

The availability of the human genome sequence has revolutionized human genetics research. By studying the differences in the genomes between sick and healthy individuals one can associate particular differences with the disease. Such association is the first step towards understanding disease cause, diagnostics and eventually therapy. In this thesis we study several problems related to the most common type of sequence differences, single nucleotide polymorphisms (SNPs). We first explore tag SNP selection criteria. We compare two algorithms that use different criteria for selecting tag SNPs: the r criterion and the prediction accuracy criterion. By performing extensive simulations with real haplotypes we show that tags selected according to the prediction accuracy criterion provide higher power to association studies. We show that the magnitude of the advantage in power is dependent on the tag density. When choosing tags at high density, both methods provide comparable and very high power, but as the tag density decreases, the advantage of the prediction accuracy criterion grows larger. We next describe a software package for genotype analysis that we developed. The software package combines state-of-the-art algorithms for genotype phasing, tag SNP selection, and association testing, with convenient visualizations. By streamlining the application of the algorithms, which were only available as batch executables previously, we make the algorithms accessible to the broad community of researchers in genetics. Lastly, we describe three studies on Crohn’s Disease performed in collaboration with gastroenterologists fromWolfson and Rambammedical centers. In these studies we applied common methods for statistical analysis and developed necessary ad-hoc adjustments, in order to identify genotype-phenotype and phenotypephenotype associations. The studies and methods used are briefly summarized here while full information is given in articles published in the literature.