|
Activity Number:
|
421
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #310284 |
|
Title:
|
Two-Way Clustering of DNA Microarray Data Using the Bayesian Plaid Model
|
|
Author(s):
|
Jiajun Gu*+ and Jun Gu
|
|
Companies:
|
Harvard University and Harvard University
|
|
Address:
|
36 Oxford Street, Cambridge, MA, 02138,
|
|
Keywords:
|
Bayesian plaid model ; bi-clustering ; normalization of DNA microarray data ; gibbs sampling
|
|
Abstract:
|
In recent years, two-way clustering or bi-clustering algorithms are in particular interest for DNA microarray analysis. A bi-cluster refers to a subset of genes and a corresponding subset of experimental conditions of a microarray dataset. In this work, a Bayesian bi-clustering model is proposed. Similar to the plaid model of Lazzeroni and Owen in 2002, the Bayesian plaid model uses an ANOVA model to represent a bi-cluster and allows an unknown number of bi-clusters. Gibbs sampling algorithm is used to search for multiple bi-clusters simultaneously. We applied the model to both simulated and real datasets including leukemia and yeast microarray data and identified bi-clusters with significant biological meanings. We also compared it with a few published bi-clustering algorithms based on their operation characteristics such as specificities, sensitivities and overlapping rates.
|
