| Summary: | 碩士 === 慈濟大學 === 醫學資訊研究所 === 94 === In the cell cycle, the gene repression or promotion is the key to the cell division and growth normally. Cancers, the number one cause of people death, are due to the abnormal growth of cell. through the study of regulatory mechanism of cell cycle, therefore, abnormal chromosome change in the cell and its relationships with cancer.
The purpose of this thesis is to develop structure learning algorithms of Bayesian network and to apply these algorithms to construct the gene network from the experimental microarray data. We intend to predict a relation between genes precisely and to express a relation between genes clearly througth the graphical model of Bayesian network. We develop various algorithms from dependent-based PC algorithm, then searching and scoring HC algorithm , MWST algorithm, K2 algorithm to MWST_K2 algorithm in order to get a most accurate metod. First, we apply our algorithms to the ALARM and Asia datasets to test how accurate our model can get. After using datasets of ALARM and Asia evaluate its accuracy and running speed, we apply these algorithms to yeast mocroarray dataset to study feasibility and performance of every algorithm further. Through the improvement of various algorithms, we find that the K2 algorithm has the best accurancy and MWST and MWST_K2 algorithms have better running performance.
|
|---|
