Tree-based Methods for Analyzing Tissue Microarray Data

• Activity Number: 61
• Type: Contributed
• Date/Time: Sunday, August 3, 2003 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xueli Liu*+ and Yunda Huang and Steve Horvath
• Companies: UCLA Healthcare Resources and University of California, Los Angeles and University of California, Los Angeles
• Keywords: tissue microarrays ; random survival forests ; survival trees ; Cox regression analysis ; C-index

Abstract

Tissue microarrays (TMAs) are a new high-throughput tool for the study of protein expression patterns in tissues and are increasingly used to evaluate the diagnostic, prognostic importance of tumor biomarkers. TMA data are rather challenging. Covariates are either ordinal variables (range 0-4) or highly skewed percentages. Tumor marker expression values are usually highly correlated. Here we propose a standard data analysis flow chart for analyzing tissue microarray data. Since it is standard practice in the TMA community to use cut-off values for tumor marker expression values, it is natural to apply tree-based methods. Specifically, we propose the following analysis steps. First, use random survival forests (Breiman 2001, Huang and Horvath 2002) to screen out "important" tumor markers. Second, use survival trees to arrive at appropriate cut-off values and dichotomize the covariates. Third, conduct a standard Cox regression analysis. Fourth, assess discriminatory power with a generalization of the area under the ROC curve to survival outcomes: the C-index (Harrel 1996). We also discuss how to incorporate pathology information into the analysis.