Bayesian methodology for genetics of complex diseases

摘要: Genetic research of complex diseases is a challenging, but exciting, area research. The early development the was limited, however, until completion Human Genome and HapMap projects, along with reduction in cost genotyping, which paves way for understanding genetic composition diseases. In this thesis, we focus on statistical methods two aspects research: phenotype definition etiology identifying potentially associated Single Nucleotide Polymorphisms (SNPs) SNP-SNP interactions. With regard to etiology, firstly investigated effects different phenotyping approaches subsequent analysis. light findings, difficulties validating estimated phenotype, proposed reconciling phenotypes models using Bayesian model averaging as coherent mechanism accounting uncertainty. second part turned SNPs SNP We review use logistic regression variable selection identification extended detecting interaction population based case-control studies. study, also develop machine learning algorithm cope large scale data analysis, namely modified Logic Regression Program (MLR-GEP), then compared model, Random Forests other variants logic regression.