MatrixClust

MatrixClust performs fuzzy clustering of symmetric weighted adjacency matrices and simultaneous clustering of genes and experimental conditions to identify coexpressed gene modules and coregulatory relationships in gene expression data.

Key Features:

• Model-Based Clustering Approach: Employs a model-based clustering algorithm adapted for simultaneous clustering of both genes and experimental conditions.
• Bayesian Framework with Gibbs Sampling: Uses a Bayesian approach with Gibbs sampling to iteratively update cluster assignments for genes and conditions.
• Support for Symmetric Weighted Adjacency Matrices: Operates on symmetric matrices represented as weighted adjacency matrices of undirected networks.
• Handling Large-Scale Data Sets: Designed to manage large microarray gene expression data sets and to identify strongly peaked posterior distributions concentrated on a limited number of equiprobable clusterings.
• Graph Spectral Method for Fuzzy Clustering: Applies a graph spectral method to extract fuzzy, overlapping clusters, identifying tight cluster cores and partial coexpression in overlapping regions.
• Biological Validation via GO Annotation Analysis: Assesses biological significance of local maxima through Gene Ontology (GO) annotation analysis.
• Application to Saccharomyces cerevisiae Data: Has been applied to three large compendia of Saccharomyces cerevisiae gene expression data.

Scientific Applications:

• Systems Biology: Identification of coregulatory modules and exploration of gene regulatory networks.
• Genomics and Gene Function Analysis: Characterization of gene coexpression patterns and functional modules in genomics studies.
• Microarray Gene Expression Analysis: Analysis of large microarray expression compendia to detect coexpressed gene sets.
• Yeast Regulatory Network Studies: Investigation of regulatory relationships in Saccharomyces cerevisiae.

Methodology:

Uses model-based clustering adapted for simultaneous clustering of genes and conditions; a Bayesian framework with Gibbs sampling to iteratively update cluster assignments; graph spectral methods on weighted adjacency matrices to extract fuzzy, overlapping clusters; and Gene Ontology (GO) annotation analysis for validation.