ClusterJudge

ClusterJudge evaluates the biological relevance of gene expression–based clustering results by measuring the association between cluster assignments and known gene annotations.

Key Features:

• Annotation-Based Cluster Evaluation: Quantifies the biological quality of clustering results using curated gene attributes such as functional categories.
• Mutual Information Metric: Computes a figure of merit based on the mutual information between cluster membership and gene annotations.
• Clustering Method Comparison: Enables systematic evaluation of clustering algorithms, distance measures, and parameter settings applied to gene expression data.
• Resolution Assessment: Assesses the relationship between the number of clusters and biological enrichment across expression datasets.
• Distance Metric Evaluation: Supports comparison of dissimilarity measures such as Euclidean distance and Pearson correlation distance for clustering performance.

Scientific Applications:

• Gene Expression Clustering Evaluation: Assesses the biological validity of clusters derived from gene expression datasets.
• Algorithm Benchmarking: Compares clustering algorithms including hierarchical clustering and self-organizing maps (SOMs) using annotation-informed metrics.
• Functional Genomics Analysis: Evaluates whether clusters correspond to biologically meaningful gene functional categories.

Methodology:

ClusterJudge calculates a figure of merit based on the mutual information between cluster assignments and curated gene attributes, enabling statistical evaluation of clustering algorithms, distance measures, and parameter configurations.