Classification and Class Discovery of Tumor Samples using Mixture
Modeling of Gene Expression Data
Dr. Roxana Alexandridis,
Department of Bacteriology, UW Madison
Friday, October 14, 2005, 12:00 noon
Classification and Class Discovery of Tumor Samples using Mixture Modeling of Gene Expression Data
Accurate classification of tumor samples is an essential
tool in cancer diagnosis and treatment. DNA microarray technology has been increasingly used in cancer research. Most of the thousands of genes whose expression levels are measured do not contribute to the separation between types of tumors. We describe several methods that address the issue of gene selection, and present a new method for tumor classification based on microarray gene expression levels. In the literature, discovery of putative classes and classification to known classes based on gene expression data have been largely treated as separate problems. We offer a unified approach to class discovery and classification, which has greater applicability in practical situations. The method we propose is based on modeling the distribution of the gene expression profile of a tumor sample as a mixture of an unknown number of distributions, each characterizing the gene expression levels in a class. We demonstrate applications to a leukemia dataset and a colon cancer dataset.