BHC

BHC performs Bayesian agglomerative hierarchical clustering of gene expression microarray data to identify probabilistic gene clusters and model uncertainty in cluster formation.

Key Features:

• Bayesian Agglomerative Hierarchical Clustering: Implements a bottom-up hierarchical clustering strategy that uses a probabilistic framework to account for uncertainty in cluster formation and merging decisions, suitable for large-scale gene expression data.
• Dirichlet Process Modeling: Uses a Dirichlet Process infinite-mixture model to allow flexible determination of the number of clusters without pre-specification.
• Bayesian Model Selection: Applies Bayesian model selection at each merging step to choose statistically justified cluster merges.
• Handling of Multinomial and Time-Series Data: Supports analysis of multinomial (discrete categorical) data and time-series gene expression datasets.

Scientific Applications:

• Gene expression microarray analysis: Clusters microarray gene expression profiles to reveal coherent expression patterns across conditions or samples.
• Arabidopsis thaliana stress-response studies: Has been applied to analyze gene expression in Arabidopsis thaliana under various stress conditions to identify groups of co-expressed genes.

Methodology:

Iterative Bayesian agglomerative merging of data points guided by Bayesian model selection, with a Dirichlet Process infinite-mixture model enabling adaptive determination of the number of clusters.