Systems Biology Analysis On Microarray Data

Systems biology is a new discipline emerged in the last two decades. It observes and studies the biological systems from a new perspective. Being different from tra-ditional molecular biology which is based on reductionism, the objective of systems biology is to understand the properties, structure and behavior of biological systems from holism view, by integrating different levels of biological data. The development of systems biology relies on massive data processing and high-throughput technolo-gies, but also relies on multidisciplinary support such as biology, computer science, statistics and physics. In this dissertation Focusing on gene regulation, we developed several new systems biology methods based on microarray data, and tried to analyze gene regulatory network which is involved in liver caner generation. The research is divided into three sections in detail:1. SiGPAT gene set analysis method. Gene expression in biological system shows characteristic of complexity and changeability, always accompanied by a wide range of random signals. The objective is to identify significant biological function-s from complex gene expression data. Understanding of these features can help re-searchers to focus on the most important category of genes and further to make close study on their mechanism. Gene set enrichment analysis is a new category of methods of microarray data analysis. It focuses on a list of biological related genes and thus has the advantage of finding out the molecule pathways that are responsible for the change of the state of the cell and biological functions which are significantly affected. In this dissertation, first, the methodology of gene set enrichment analysis was discussed in detail. We described various analytical frameworks in a modular fashion. Then, for the limit of current gene set enrichment analysis methods, we proposed a new model, aim-ing to find gene sets having significant expression patterns. In the model, two statistics are designed which are the up-regulation potential and down-regulation potential of the gene set to explain how genes in the gene set are differently regulated. Through the simulation experiment, it shows that our model can correctly identify gene sets with different expression patterns. In the analysis of two real-world microarray data sets, it shows that our model can reveal biological related results and improve the robustness in microarray data analysis.2. CePa pathway enrichment method. For a special class of gene sets, biolog-ical pathways contain more important information that is the complicated interactions among biological molecules. Biological pathways are a set of genes or molecules that act together in form of chemical reactions, molecule modifications or signaling transduction to carry out biological functions. We proposed a new model that extends current pathway enrichment methodology, by integrating the network topological in-formation of pathways. Our model can highlight the status of key genes in the network of pathways. In addition, to be consistent to the real biological situation, we use net-work node as the basic unit for analysis, instead of gene. This is because in the real case, the gene must be assembled into complexes to function normally. As long as one of members is disordered, the function of the entire complex will be severely affected. Through a simulation study, we showed how different network structures and different centrality measurements affect the significance of pathways. In the analysis of liver cancer data set, our model was successful to find relevant biological process which cannot be found by previous method, and we demonstrated how key genes affect the significance of the pathways that directly relate to biological meanings.3. Construction and systems biology analysis of gene regulatory network in liver cancer. We attempted to reveal the most important mechanism of gene regulatory network from a systems perspective. We reconstructed gene regulatory network that is involved in liver cancer generation. The regulators in the regulatory network contain both transcription factors and microRNAs. The interactions were integrated from three data sources, which are target prediction, experimentally-supported interactions and microarray data. Three data sources assign relations to regulators and targets from different aspects, and the integration of the three data sources can make the result more reliable. Due to the different mechanisms of regulation of transcription factors and microRNAs, parameters for the two types of regulators are selected and optimized separately. Analysis of the gene regulatory network revealed that gene regulation in liver cancer is highly modular, in which microRNAs mainly regulate functions related to mitochondria and oxidative reduction, and transcription factors regulate functions related to immune response, extracellular activity and cell cycle. On the higher level of gene regulation, there exists a core network that regulates different modules. The core network was critically important in maintaining the stability and robustness of the network. The core network proposed is highly related to liver cancer, and we believe it reflects the key mechanism of gene regulation in liver cancer, and serve as the basis for the downstream experimental biology research.In this dissertation, we started from revealing the basic properties of gene regula-tion system, and then extended to the local structure of the system; finally we studied the overall structure of the biological system under specific disease. We have devel-oped several efficient models and analysis workflow, succeeded in analysis of gene regulatory network in liver cancer. With the popularity of high-throughput gene test-ing technology, the result of this study will play extensive role in disease mechanism research and new drug finding.