Construction and evaluation of yeast expression networks by database-guided predictions

Area of Science:

• Molecular Biology
• Systems Biology
• Bioinformatics

Background:

• DNA microarrays provide genome-wide expression data.
• Toxic polyglutamine proteins induce changes in yeast gene expression.
• Analyzing these changes requires understanding transcriptional networks.

Purpose of the Study:

• To develop and evaluate an algorithm for constructing transcriptional networks from yeast microarray data.
• To improve the statistical capabilities and predictive power of network analysis.
• To facilitate the assignment of biological processes and transcription factors to gene clusters.

Main Methods:

• Developed a clustering algorithm using co-regulatory relationships from the SPELL database.
• Evaluated algorithm performance based on network quality and predictive accuracy.
• Implemented a scoring method to quantify network predictive strength.
• Tested the algorithm on experimental and public microarray datasets.

Main Results:

• The algorithm effectively constructs networks even with limited microarray data.
• Network quality is influenced by the number of co-regulated genes and input hits.
• Predicted genes improve network quality and allow for quantitative assessment of predictive strength.
• Clusters significantly enhance the assignment of biological functions and transcription factors.

Conclusions:

• The developed clustering approach and evaluation parameters are valuable tools for analyzing yeast microarray data.
• The ClusterEx web interface will provide access to this method.
• This approach enables fast and informative interpretation of gene expression changes.