Prediction of Cancer Drug Sensitivity Using High-Dimensional Genomic Features

• Activity Number: 224
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Ting-Huei Chen*+ and Wei Sun
• Companies: and University of North Carolina at Chapel Hill
• Keywords: cancer drugs ; drug sensitivity

Abstract

A large number of cancer drugs have been developed to target genes/pathways that are crucial for cancer growth. Several drugs that share a target may also have some common predictive genomic features. Therefore, it is desirable to analyze these drugs as a group to identify the associated genomic features. Furthermore, these genomic features may be robust predictors for any drug sharing the same target. The high dimensionality and the strong correlations among the genomic features are the main challenges of this task. Motivated by this problem, we develop a new method for high-dimensional bi-level feature selection using a group of responses that may share a common set of predictors in addition to their individual predictors. Simulation results show that our method has a substantially higher sensitivity and specificity than existing methods. We apply our method to two large-scale drug sensitivity studies. Within-study cross-validation demonstrates that the genomic features identified by our method have high prediction power. Between-study validation shows that the genomic features selected for a drug target can form good predictors for other drugs designed for the same target.