A New Support Vector Regression Approach to Gene Selection

• Activity Number: 350
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2008 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Pei-Chun Chen*+ and Su-Yun Huang and Chuhsing K. Hsiao+
• Companies: National Taiwan University and Institute of Statistical Science, Acadamia Sinica and National Taiwan University
• Keywords: SNP ; support vector machine ; support vector regression

Abstract

Genetic data are often collected with a large number of genes and a relatively small simple size. However, only a few genes among them are associated with the disease of interest. As the number of markers increases, the statistical analysis becomes complicate due to the dimensionality. A great challenge is to find the genes showing significant association with the disease. Most approaches focus on a multi-hypotheses testing of the gene-disease association. Recently, the statistical machine learning methods are applied to this problem, such as support vector machine (SVM). We propose a new approach for selection by weighting on every sample differently. These weights are calculated via support vector regression. A SNPs data set of schizophrenia will be analyzed for illustration.