Biomarker Identification and Patient Condition Prediction Based on Gene-Expression Pattern from Two Case Studies

• Activity Number: 610
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yushi Liu*+ and Joe Verducci
• Companies: The Lovelace Respiratory Research Institute and The Ohio State University
• Keywords: SCOOP ; Microarray ; Biomarker ; cancer ; SVM

Abstract

In this study, we used different methods to select biomarkers showing distinctive gene expression profiles between patients with breast cancer and normal people based on the microarray gene expression data. Different from the original methods, we used SVM (Supporting Vector Machine) based on the biomarkers selected by different methods to classify the cancer status of the new patients. We found SVM combined with the gene filtering techniques resulted in higher prediction accuracy than the author claimed. As an extension, we also used similar to classify small, round blue cell tumors into four different subcategories based on the study of gene expression profile. Since SVM was only suitable for classification into two category. The structured SVM served as an extension of the regular SVM in this study. And combination of the structured SVM and the gene filtering techniques resulted in mu