November 18, 2006
We developed ExpressionNet, a program that uses Bayesian network learning algorithms to explore relationships among random variables to generate network models that best explains external data. The software has been used to study the transcriptional response in budding yeast. A combination of gene expression clusters, genetic information and experimental conditions was used to derive network that models gene expression response to a seriese of environmental triggers. Detail of the study has been submitted for publication.
The program has been applied to model the transcriptional response to nitric oxide in Saccharomyces cerevisiae, which will soon be published. The following is the supplemental data for the manuscript.
Jingchun Zhu, Ashwini Jambhekar, Aaron Sarver and Joseph DeRisi
The supplemental data are organized by the three iterations of modeling. Network models (files with extension .bn or .matrix) can be viewed using ExpressionNet, a free program developed for Bayesian network learning, which is available at http://expressionnet.sourceforge.net. Microarray clustering results (files with extension .cdt or .gtr) can be viewed as a pair using JavaTreeView, a free program available at http://jtreeview.sourceforge.net.
The datasets include the derived network models, the learning datasets, gene cluster memberships, and complete microarray datasets obtained in this study.
Networks used to construct the model: in the “1st model high-scoring network collection” folder
High-scoring criteria: top 15 percentile using both priors
Networks used to construct the model: “2nd model high-scoring network collection” folder
High-scoring criteria: top 15 percentile using two both priors
Networks used to construct the model: “3rd model high-scoring network collection” folder
High-scoring criteria: the top 25 percentile using both priors