Abstract
Understanding of gene regulation provides biologists applied in many important drug study including drug target identification and drug development. However, gene regulatory relationships are not yet well understood now. Although Microarray technology provides a large-scale measurement of gene expressions, it also contains large missing gene expressions because of noisy. How to infer correct gene regulatory networks from microarray data is a big challenge to biologists. In this thesis, we present a workflow for inferring gene regulatory networks from micorarray data based on transcription factor analysis and the conditional independence. The system that constructs gene regulatory networks consists of three components: (I) Microarray data preprocessing, (II) Biological knowledge processing, and (III) Revising and inferring process. We reconstruct the gene regulatory networks using d-separate criteria and conditional independency in Bayesian Network. And then, we use rules to infer interactions (activation or inhibition) of links in the gene regulatory network. The system also integrates the bioinformatics toolkits and databases to automatically extracts the promoter regions of DNA sequences to predict the transcription factors that regulate the gene expressions. We visualize the reconstructed gene regulatory network that the analyzed microarray results are easy to understand. We analyze two microarray datasets from Stanford Microarray Database. We use two approaches to evaluate our result: (1) the topology of network and (2) verifications based on literature reports. The evaluated results show that the gene regulatory networks we reconstructed could support the process of detecting possible cancer genes.