Dissertation > Excellent graduate degree dissertation topics show

Research on Microarray Gene Expression Data Based on Bayesian Network

Author: ZhangYan
Tutor: DaiDaoQing
School: Sun Yat-sen University
Course: Information Computational Science
Keywords: Gene Bayesian networks Posterior probability Structure Learning Parameter learning
CLC: Q78
Type: Master's thesis
Year: 2010
Downloads: 71
Quote: 0
Read: Download Dissertation

Abstract

With the completion of the draft sequence of the human genome, the functional genome research to occupy an increasingly important position in the field of life sciences. Elucidate the molecular mechanisms dependent regulation of gene selectively expressed their interactions, reveals the core of the essence of the phenomenon of life, functional group research. With the the genomics research expanded in-depth, the regulation of gene expression from a single gene, the linear regulation to expand to the stereo level multi-gene, gene cluster and the whole genome of regulatory networks. How to effectively use existing genomics data, fully integrated multi-disciplinary ideas, create a new test system and the technical system to clarify the regulation of gene expression by network, analysis of mutual restraint relationship between the gene has become the field of functional genomics within the focus of international competition. Bayesian network approach to knowledge of probability theory and graph theory, and its graphical representation, clear causality uncertainty reasoning advantages Bayesian network to microarray data analysis, probability angle describes the dependence relation between the respective genes, thus clarified between the regulatory network in the entire genome. This paper describes the basic concepts of Bayesian networks, the development of history and classification of its characteristics, as well as some of the basic methods of Bayesian network structure, elaborated Bayesian network structure learning and parameter learning principle, and then sub- several cases the structure learning and parameter learning combine to create a Bayesian network model of the interaction between genes. In the case study, this article describes in detail the entire process of microarray data Bayesian network model. In this paper, the gene expression data to build a three-valued discretization Bayesian network, and analysis of the role and influence of several genes with more child nodes in the network. This paper was the study of the following aspects: (1) in the complete data set, use the K2 Bayesian network structure learning algorithm for structure learning research, K2 need to determine in advance the sort between each node, this when using the decision tree algorithm to complete the scheduling problem, thereby improving the efficiency of learning. And set the number of the largest parent node in the network learning process. Eventually obtained through structured learning Bayesian network model, based on the use of the maximum likelihood method to estimate parameters, thus capturing the posterior probability of genes may reflect the regulatory relationships between genes. Last Bayesian equivalence classes is obtained, found in the original network can be inverted to the arc, which can be carried by means of experiment or the like to obtain a priori knowledge corresponding adjustment of the network structure, doing so does not affect the structure of the network , more practical. (2) on the basis of prior knowledge that the network structure is known, random data set of 1/3, 1/4, 1/5 of the missing values ??using the expectation-maximization algorithm parameters to learn, to obtain the desired maximum processing capabilities of the algorithm on data with missing values. (3) in the absence of a priori knowledge contains missing data on the basis of structural expectation maximization algorithm complete learning the network structure and the corresponding parameter structure and get the structure expectation maximization (EM) algorithm processing power. Bayesian networks have good theoretical knowledge and a clear form of knowledge representation is the uncertainty study an important way, has an important role in data mining. Introduced into the analysis of microarray experimental data can be better to build a network model to analyze interactions between gene and influence can be widely applied to the study of biology and oncology, observe the diseases caused by changes in gene expression, and find out the important role of the variable genes.

Related Dissertations

CLC: > Biological Sciences > Molecular Biology > Genetic engineering (genetic engineering)
© 2012 www.DissertationTopic.Net Mobile