Cancer Diagnosis By Microarray And Bioinformatics Methods

With the advent of Human Genome Project launched by the United States in the 1990s, data explosion that has never emerged before became the research focus of thousands of scientists. As far as a great number of materials produced by the biological experiments are concerned, original data is the source of information and beginning of knowledge. More importantly key for life and nature can be probably reached in the exploration of information and knowledge. Therefore Bioinformatics comes into being, which has been significantly motivated by this great incentive. Bioinformatics has become a powerful tool to reveal the biological meaning by a series of data-oriented work, including data acquirement, data modification, data storage and data analysis.Meanwhile Bioinformatics has played an important role in the study of functional gene identification, illness diagnosis, gene expression profiling and protein function. Especially in virtue of microarray technology and data mining work different malign tumors with similar clinical responses and histopathological appearance can be precisely and promptly classified. It is well known that cancer treatment is largely influenced by the tumor development phase. In another word, the earlier the cancer is diagnosed, the better therapy effect can be gained for the patients. To the great extent integration of microarray technology and Bioinformatics methods devotes much to the early diagnosis and timely treatment of cancer.In this paper, pattern recognition methods are main research work, which primarily are used for discriminating the different kinds of cancer by gene expression profiling produced by microarray technology. Considering that complexity of array data and advantage of different classifiers, two classification systems are employed. However both of them follow the same fundamental process: first dimensionality reduction, then modeling and finally prediction. For the first way gene selection and data reconstruction based on kernel trick are separately adopted for dimensionality reduction and the classifier is Self-Organizing Map, one of the most popular Artificial...