Statistical Methods For Haplotype Analysis With Genotyping Errors

Haplotype plays a very important role in modern genetic epidemiology studies. Especially, in the study of mapping common complex disease genes, haplotype-based methods, such as linkage analysis and association analysis, are more powerful than single SNP marker methods. In practice, however, what we can obtained directly are genotype data but not haplotype data. The basic problem of haplotype analysis is that we should infer haplotypes of each individual according to the available genotype data, then we should perform haplotype-based linkage analysis and haplotype-based association analysis. Nevertheless, almost all of the existing haplotype analysis (haplotype inference, haplotype-based linkage analysis and haplotype-based association analysis) neglect the impact induced by genotyping errors, which perform haplotype analysis under the assumption that genotype data do not contain errors. The major issue is that all large genotype data, especially for SNP markers, contain errors due to fallible genotyping technologies. This dissertation aims to propose some new statistical methods for haplotype analysis with genotyping errors.In this dissertation, we present several haplotype inference methods for population data and pedigree data respectively when genotype data contain some errors, furthermore, we present a new haplotype association method to reduce the impact induced by genotyping errors. In the study of haplotype inference for population data, two novel strategies, double sampling strategy and multi-genotyping strategy, are proposed for constructing "GenoSpectrum" of each individual, then according to the new strategies, two algorithms, DS-EM and MG-EM, are proposed for haplotype inference with genotyping errors. For pedigree data, we present a GS-PEM algorithm to perform haplotype inference with geno-...