Benefiting from high throughput technologies, significant progress has been made by genome wide association studies (GWASs) and gene expression profiling to map genetic susceptibility region for complex human diseases. Evidences show that the current findings can only explain part of the genetic etiologies, which could be partially due to the noise and artifacts introduced by high throughput technologies and the deficiencies in the current available analytical tools. To reduce the effects of noise and artifacts and facilitate the genetic studies, I develop three statistical methods which aim at 1) providing accurate genotype calls by modeling the underlying hybridization process of microarray with the consideration of batch effect; 2) reducing false positive and false negative findings for differentially expressed gene identification by incorporating the variability of data preprocessing into the differentially expressed gene detection algorithm and 3) studying the genetic etiologies contributing to comorbidity between complex human diseases by proposing a multivariate Mann-Whitney method built based upon a U-statistic with forward selection algorithm. Through simulations, analyses of the Latin Square Data, and the HapMap data, I show that the three proposed methods outperform the current existing methods and are robust under various experimental conditions and disease models. I further apply these methods to datasets obtained from Wellcome Trust Case Control Consortium to identify the genetic susceptibility loci predisposing to coronary heart disease and to the comorbidity between coronary heart disease and Type II diabetes. With these newly developed methods, the loci identified for coronary heart disease are consistent with the findings by various technologies, which indicates the proposed method could provide accurate genotype calls and benefit the downstream analysis. No loci have been selected to be associated with the co-morbidity of coronary heart disease and type II diabetes which may be due to the study design and the candidate gene approach used in this research. Further studies are needed to investigate the comorbidity between coronary heart disease and type II diabetes. I also apply my method to identify differentially expressed genes for a cervical cancer study. The findings replicate most of the original discoveries. In addition, several other genes, which potentially play an important role in the cervical cancer development, have also been identified.
Read
