Using Gene Gxpression to Predict Dose-Response Data

• Activity Number: 406
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Philippe Haldermans*+ and Ziv Shkedy
• Companies: I-Biostat and Universiteit Hasselt
• Keywords: microarray ; dose-response data ; prediction method ; lasso

Abstract

DNA microarrays allow us to monitor thousands of genes simultaneously. One area of interest is oncology, where we want to predict the response of a patient to a certain compound, using the microarray data. The burden of microarrays is the fact that the number of variables (p) is much larger than the number of subjects (n). We compare existing large scale prediction methods, such as Supervised Principal Component Analysis and Lasso, to the new method we propose called Weighted Ensemble Prediction. The existing prediction methods might lose accuracy when the number of genes increases to tens of thousands. The goal of our method is to remove uninformative genes before applying the prediction method. Similarly to Enriched Random Forest, ensembles of genes are used to identify important genes. We use both simulated data and real datasets to compare our method to existing methods.