Chapter 6 - Statistical Analysis of High-Dimensional Data for Pancreatic Cancer

Abstract

Pancreatic cancer is a highly malignant disease with a five-year survival rate less than 5%. Advances in modern data collection technology have revolutionized the way that we study the complex biological systems, allowing pancreatic cancer researchers to make genome-wide expression profiling within tumors in a fast, precise, and cost-effective way. How to correctly analyze and interpret the high-dimensional and complex gene expression data is a key to understanding the hidden regulatory mechanisms. In this work, we first introduce a LASSO penalized Cox regression method to identify individual genes that are directly related to survival time of pancreatic cancer patients. A cyclic coordinate descent algorithm is used for the computation of high-dimensional data (number of genes larger than number of patients). Then, we introduce a doubly regularized Cox regression method, which integrates pathway information into our analysis, to identify both genes and signaling pathways related to pancreatic cancer survival. Both methods are applied to a pancreatic cancer microarray dataset and identify several genes and signaling pathways correlated to pancreatic cancer survival. Our findings can help cancer researchers design new strategies for the early detection and diagnosis of pancreatic cancer.