Coupling a Frequency-Based Feature Selection Method and Voting Classifier Approach for Biomarker Detection

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Sandra L. Taylor and Kyoungmi Kim*+
• Companies: University of California at Davis and University of California at Davis
• Keywords: Biomarker detection ; feature selection ; classification ; omics

Abstract

With technological advances, the -omics fields have recently explored and applied to develop diagnostic and prognostic tests. An investigator's objective is to develop a classification rule to predict class of unknown samples based on a small set of features that can ultimately be used as biomarkers in a clinical setting. A number of methods have been developed for feature selection and classification using gene expression data, proteomics or metabolomics data. However, the great challenge is not the design of algorithms but the potential application in a clinical setting. While common classification methods such as random forest (RF) and support vector machines (SVM) are effective at separating groups, they do not directly translate into a clinically-applicable classification rule based on a small number of features. In this study, we present a frequency-based feature selection method coupled with a voting classifier approach that yields "stable" feature sets. We evaluate the performance of voting classifiers using three -omics datasets. We show our approach achieves classification accuracy comparable to RF and SVM while yielding classifiers with clear clinical applicability.