|
Activity Number:
|
63
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #310340 |
|
Title:
|
An Efficient Mixture Model Approach To Characterize Gene Pathways Using Bayesian Networks
|
|
Author(s):
|
Younhee Ko*+ and Sandra Rodriguez-Zas and Chengxiang Zhai
|
|
Companies:
|
University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign
|
|
Address:
|
1207 W Gregory Dr, Urbana, IL, 61801,
|
|
Keywords:
|
microarray ; mixture of models ; Bayesian Information Criterion ; Bayesian networks
|
|
Abstract:
|
Traditionally, the identification of directional gene networks using Bayesian networks requires the discretization of continuous gene expression measurements that may result in loss of information. A weighted mixture of Gaussian models was used in a Bayesian network to identify gene pathways from continuous gene expression data. Parameter estimates were obtained using an EM algorithm and the network structure better supported by the data was identified using Bayesian Information criterion. To overcome high computational demands, the potential number of genes influencing the expression of another gene was restricted without jeopardizing the parameter search space. The Bayesian network was applied to two published yeast pathways and a honey bee pathway. Results indicated that our approach can efficiently characterize gene pathways with different topology using continuous data.
|
