Searches for the elusive genetic mechanisms underlying complex diseases have long challenged human geneticists. Recently, genome-wide association studies (GWAS) have successfully identified many complex disease susceptibility loci by genotyping a subset of several hundred thousand common genetic variants across many individuals. With the rapid deployment of next-generation sequencing technologies, it is anticipated that future genetic association studies will be able to more comprehensively survey genetic variation, both to identify new loci that were missed in the original round of genome-wide association studies and to finely characterize the contributions of identified loci. GWAS, whether in the current genotyping-based form or in the anticipated sequencing-based form, pose a range of computational and analytical challenges.

I first propose and implement a computationally efficient hidden Markov model that can rapidly reconstruct the two chromosomes carried by each individual in a study. To achieve this goal, the methods combine partial genotype or sequence data for each individual with additional information on additional individuals. Comparisons with standard haplotypers in both simulated and real datasets show that the proposed method is at least comparable and more computational efficient.

I next extend my method for imputing genotypes at untyped SNP loci. Specifically, I consider how my approach can be used to assess several million common variants that are not directly genotyped in a typical association study but for which data are available in public databases. I describe how the extended method performs in a wide range of simulated and real settings.

Finally, I consider how low-depth shot-gun resequencing data on a large number of individuals can be combined to provide accurate estimates of individual sequences. This approach should speed up the advent of large-scale genome resequencing studies and facilitate the identification of rare variants that contribute to disease susceptibility and that cannot be adequately assessed with current genotyping-based GWAS approaches.

My methods are flexible enough to accommodate phased haplotype data, genotype data, or re-sequencing data as input and can utilize public resources such as the HapMap consortium and the 1000 Genomes Project that now include data on several million genetic variants typed on hundreds of individuals.