A Study On Identifying Gene Markers For Ovarian Carcinoma Chemotherapy Prediction

Ovarian cancer is a common and the most lethal type of gynecological cancer. Primary debulking surgery followed by chemotherapy is currently the standard of care for patients with advanced ovarian cancer. However, patients have differential responses to chemotherapy, predicting the prognosis of chemotherapy therefore become paramount important. Recent advances inthe use of DNA microarrays, which allow global assessment of gene expression in a single sample, have shown that expression profiles can provide molecular phenotyping to identify chemotherapy responses.So taking OvCa DNA microarray expression data analysis as the research topic, this dissertation refers to studies on relative preprocessing techniques and gene feture selection methods. The main contents and creative contributions of the dissertation are summarized as follows:(1) Microarray expression data preprocessing method. Microarray data carried out in different labs can cause gene different expression. To this end, a preprocessing method is proposed, which shows non-biological causes of gene different expression in the way of graph and statictics. Results show that genes different expression has no such non-biological causes after being preprocessed.(2) A wrapper and filter combined approach is proposed. With the advancement of microarray technology, many gene markers selection approaches have been proposed for cancer diagnosis. However the factor of the biological relevance is not efficiently incorporated into either the filter or the wrapper based gene selection methods. To this end, we propose to identify biomarkers primarily in terms of their chemotherapy responses, and then adjust the biomarkers based on the classification quality. A wrapper and filter combined approach was proposed. Compared to other existing methods, this approach gained higher accuracies and more robust to noise. The identified biomarkers demonstrate closer relationship with the ovarian carcinoma chemotherapy and the biological relevancies are supported by independent research as well.(3) Some new biomakers are identified, which show closer relationship with the ovarian carcinoma chemotherapy. We identified 45 biomarkers demonstrate close relationship with the ovarian carcinoma chemotherapy in many aspects. Specifically, among them, there are 4 genes NR2F2, CLDN3, PURA, C1ORF38 which are directly related to ovarian carcinoma; 17 genes are shown to be associated with cancers, tumors, immune responses or inflammatory responses, namely, USO1, TCF7L2, NR2F2, MGEA5, CLDN3, DUSP6, ENO1, TANK, VLDLR, DDAH1, SMARCA2, C1ORF38, SLC22A5, PURA, ACTR2, MAP3K4, OGT.This study identified gene markers which are closely related to ovarian carcinoma and would greatly contribute to the prognosis and treatment of ovarian carcinoma.