Gene Expression--Based Predictive Models for Cancer Drug Sensitivity

• Activity Number: 123
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Umut Ozbek*+ and Jaya Satagopan
• Companies: Icahn School of Medicine at Mount Sinai and Memorial Sloan Kettering Cancer Center
• Keywords: Pharmacogenomics ; regularization methods ; principal components

Abstract

Pharmacogenomic studies aim to elucidate the genetic or genomic features contributing to the efficacy of medications. The analysis of these data are challenging due to the large number of genomic features that must be investigated. Shrinkage or regularization techniques can help identify the most important features that can predict sensitivity to drug. In this work, we review the properties of several regularization methods using gene expression data from the Cancer Cell Line Encyclopedia project for predicting the sensitivity of 74 lung cancer cell lines to the drug, Nilotinib. We provide an empirical illustration of the predictive performance of the regularization methods based on three approaches - the first that directly uses gene expressions, and two novel methods that use principal components of the gene expressions.