|
Activity Number:
|
372
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #303537 |
|
Title:
|
Modeling the Spatial Structure in Microarray Data
|
|
Author(s):
|
Vinicius D. Mayrink*+ and Joseph E. Lucas
|
|
Companies:
|
Duke University and Duke Institute for Genome Science and Policy
|
|
Address:
|
1500 Duke University Rd, Durham, NC, 27701-2924,
|
|
Keywords:
|
Markov Chain Monte Carlo ; Bayesian analysis ; Spatial statistics ; Microarray data ; Gene expression
|
|
Abstract:
|
A number of steps are required to convert the raw measurements in microarray technology into higher level data, whereby a single value is assigned to a gene or perhaps a splice variant. The manipulations applied to the data (preprocessing) need to account for several sources of variation, otherwise the quality of the final statistical summaries is affected. The probe level data on a single microarray are structurally organized in a lattice. Therefore, the spatial dependence between observations is an important issue to consider. Standard preprocessing techniques do not account for sources of variation related to this aspect. This study aims to use spatial information in the process of generating gene expression. The Bayesian framework is configured by employing the CAR model. The performance of the proposed Bayesian model is investigated through the analysis of simulated and real data.
|
