Modeling Spatial Structure Using High-Throughput Data

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Ke Wang*+ and Xinlei Wang and Guanghua Xiao
• Companies: Southern Methodist University and Southern Methodist University and UT Southwestern Medical Center
• Keywords: microarray ; bayesian spatial model

Abstract

High throughput technology, which allows for simultaneously collecting large amount of data, has been developed as an important tool in scientific discovery. The density and volume of the data generated in a single experiment continue to grow quickly due to advanced technology. These high density data are often spatially correlated. Since it is typical in practice that a few replicates available due to high cost, modeling the spatial correlation can improve estimation efficiency, and lead to more reliable scientific findings. We use regression model to detect differentially expressed genes in probe level. The spatial correlation of coefficients across different locations (probes in our study) can be modeled using AR(1) model. Both estimators can be smoothed for the expression level of control group, and for the expression level difference between control and treatment groups.