R Library to Validate a Gene Signature in Multiple Gene Expression Data Sets Using Principle Component Analysis

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): William James Fulp*+ and Dung-Tsa Chen
• Companies: Moffitt Cancer Center and Research Institute and Moffitt Cancer Center and Research Institute
• Keywords: Gene Signature Validation ; Microarray ; Principle Components Analysis ; R Library

Abstract

Validation of a gene signature often requires its test on independent datasets to demonstrate the association of the signature with clinical outcomes. While many public gene datasets exist, cleaning these datasets and validating a gene signature is nontrival. To quickly allow for gene signature validation, we use R software to develop libraries to collect gene expression data for a specific cancer type, and an algorithm based on principal component analysis to generate a weighted summary score to reflect the combined effect. Moreover, the library includes functions to generate various plots, such as survival curves or boxplots, depending on clinical outcome, for the weighted summary score and univariate analysis results. To date, we have created two libraries, one for breast cancer (12 datasets), and one for ovarian cancer (6 datasets). By utilizing our R libraries for gene expression data, we can test multiple new gene signatures for their clinical association in a timely fashion. We are working on building libraries for other cancer types, such as colon cancer and lung cancer.