Statistical and Computational Methods for Predicting Cancer Prognosis by Integrating Omics Data

• Activity Number: 410
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Lizhen Peng*+ and Xuefeng Wang
• Companies: Stony Brook University and Stony Brook University
• Keywords: Prediction ; Cancer Prognosis ; Genomic Measurements ; TCGA ; Feature Selection ; Dimensionality Reduction

Abstract

Thanks to the high-throughput assay technologies, various types of genomic profiles including mRNA gene expression, DNA methylation, reverse phase protein array, microRNA and copy number alteration for more than 30 cancer types are collected and publically available by The Cancer Genome Atlas (TCGA). In this work, we apply several existing methods, with integration of multidimensional genomic data, clinical information, to predict cancer prognosis. We are particularly interested in several cancers, Breast Invasive Carcinoma (BRCA), Glioblastoma Multiforme (GBM), Ovarian Serous Cystadenocarcinoma (OV) and Pancreatic Adenocarcinoma (PAAD). In order to overcome the high-dimensionality problem associated with genomic measurements, we explored most commonly adopted approaches, such as Principal Component Analysis, Partial Least Square and Regularized Least Square, and other feature selection methods, then fit the selected features with Cox survival models. We evaluate our models' performance by the predictive power for each cancer type.