Mining Gene Expression Data for Identifying Cancer Subtypes via Supervised Clustering

• Main Authors: Wen-Tzu Lee, 李文慈
• Other Authors: 陳春賢
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/35345797152213077210

碩士 === 長庚大學 === 資訊管理研究所 === 93 === With the successful decoding of human genome, it was proved that many diseases such as cancers are closely related the expression of genes. Therefore, how to increase the survival rate of cancer patients by taking advantage of gene expression data becomes one of urgent research topics. Data mining is a very popular technique for the analysis of large amount of gene expression data. Data mining technique can be used to extract potentially useful information or knowledge from complicated data sets. In this thesis, a supervised clustering method is proposed to mine gene expression data of cancer patients for the analysis of cancer subtypes. Each subtype is represented by a cluster of gene expression samples from patients of certain cancer. The clusters are further verified by using various cluster validity indices. In the verification, the optimal clustering result is derived by the ensemble of the various cluster validity indices. The proposed ensemble approach is validated by using a simulation data initially. Then gene expression data of leukemia patients are used to find the subtypes of leukemia. This thesis proposes a method for the identification of gene expression profiles of cancer subtypes. The identified subtypes of certain cancer can serve as important information to help physicians treat patients of the cancer patients affectively and appropriately.