Analysis of a Gibbs sampler method for model-based clustering of gene expression data

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Numerous clustering algorithms exist for gene expression data analysis.
• Model-based clustering offers statistical rigor but requires thorough evaluation on large datasets.

Purpose of the Study:

• To analyze the properties of a model-based clustering algorithm extended for simultaneous gene and condition clustering.
• To assess the algorithm's performance on large-scale gene expression data.

Main Methods:

• Extended an existing model-based clustering algorithm using a Bayesian approach and Gibbs sampling.
• Applied the algorithm to three large Saccharomyces cerevisiae gene expression datasets.
• Utilized Gene Ontology (GO) annotation for biological significance assessment.

Main Results:

• The algorithm effectively identifies multiple, biologically significant clusterings in large datasets.
• Simultaneous gene and condition clustering outperforms gene-only clustering.
• Fuzzy, overlapping clusters reveal complex co-expression relationships, including partial coexpression.

Conclusions:

• The developed algorithm provides robust co-clustering of genes and conditions.
• It enhances the extraction of biologically relevant information from large gene expression datasets.
• The approach facilitates the discovery of intricate gene regulatory networks.