Research Of The Method For Extracting The Tumor Gene Expression Profiles’ Informative Gene

DNA microarray technology as an important means of molecular, it monitoring on the tens of thousands of genes expression in different tissues. It not only provide strong scientific basis for the study of cancer biology, but also provides help for the identification and classification of cancer tissue. DNA microarray technology because of its high efficiency, high throughput has been widely used in biomedical research and this technology can detect gene expression of tumor massively and get its profile. We use the tumor gene expression profile to classify the tumor which has the similar organizational characteristics and to achieve early diagnosis and treatment of malignant tumor. Thus, DNA microarray technology because of its huge development potential provides an effective means for the accurate diagnosis and classification of tumors, and has the great significance in the treatment of cancer.However, due to tens of thousands of genes to be detected and higher test costs which led to the small sample. Because of the high dimensionality, high noise and small sample problems of expression profiles of tumor gene, so the traditional method of gene research are not very good to reduce the impact of these problems, can not be efficient choice for a small differential expression genes from a large number of genes. Therefore, looking for a simple and effective method to eliminate the irrelevant genes and reducing redundant genes to reduce the dimensionality of data, and finally select the optimal feature gene to improve the accuracy of classification is an important topic of this paper.Through the research about data preprocessing and data reduction, this paper for the problem of the colon cancer feature gene select use the mixed method to remove the irrelevant genes which combined filter with wind. Secondly, the Principal Component Analysis (PCA) of linear dimensionality method was applied to the processed data to reduce the effect of redundant genes on the sample classification. Finally, this article combining the feature gene selection with Support Vector Machine(SVM) classification to obtain optimal feature gene and improve classification accuracy.Using the above method, this paper studied the problem of feature extraction and samples classification, found the representative gene in the colon cancer and achieved better classification results.