Search

Nowadays, more and more high-throughput genomic data sets are publicly available; therefore, performing meta-analysis to combine results from independent studies becomes an essential approach to increase the statistical power, for example, in the detection of differentially expressed genes in microarray studies. In addition to meta-analysis, researchers also incorporate pathway or clinical information from external databases to perform integrative analysis. In this thesis, I will present three projects which encompass three types of integrative analysis. First, we perform a comprehensive comparative study to evaluate 12 microarray meta-analysis methods in simulation studies and real examples by using four quantitative criteria: detection capability, biological association, stability and robustness, and we propose a practical guideline for practitioners to choose the most appropriate meta-analysis method in real applications. Second, we develop a meta-clustering method to construct co-expressed modules from 11 major depressive disorder transcriptome datasets, incorporated with GWAS and pathway information from external databases. Third, we propose a computationally feasible algorithm to call genotypes with higher accuracy by considering family information from next generation sequencing data for two purposes: (1) to propose a new genotype calling algorithm for complex families, and (2) to extend our algorithm to incorporate external reference panels to analyze family-based sequence data with a small sample size. In conclusion, we develop several integrative methods for omics data analysis and the result improves public health significance for biomarker detection in biomedical research and provides insights to help understand the underlying disease mechanisms.

Nowadays, more and more high-throughput genomic data sets are publicly available; therefore, performing meta-analysis to combine results from independent studies becomes an essential approach to increase the statistical power, for example, in the detection of differentially expressed genes in microarray studies. In addition to meta-analysis, researchers also incorporate pathway or clinical information from external databases to perform integrative analysis. In this thesis, I will present three projects which encompass three types of integrative analysis. First, we perform a comprehensive comparative study to evaluate 12 microarray meta-analysis methods in simulation studies and real examples by using four quantitative criteria: detection capability, biological association, stability and robustness, and we propose a practical guideline for practitioners to choose the most appropriate meta-analysis method in real applications. Second, we develop a meta-clustering method to construct co-expressed modules from 11 major depressive disorder transcriptome datasets, incorporated with GWAS and pathway information from external databases. Third, we propose a computationally feasible algorithm to call genotypes with higher accuracy by considering family information from next generation sequencing data for two purposes: (1) to propose a new genotype calling algorithm for complex families, and (2) to extend our algorithm to incorporate external reference panels to analyze family-based sequence data with a small sample size. In conclusion, we develop several integrative methods for omics data analysis and the result improves public health significance for biomarker detection in biomedical research and provides insights to help understand the underlying disease mechanisms.