ClusterMine

ClusterMine integrates annotated gene sets into clustering analysis to identify sample groups and prioritize gene sets that drive sample similarity in gene expression data.

Key Features:

• Knowledge-Integrated Clustering: Partitions gene expression data into subdata based on predefined annotated gene sets to assess functional contributions to sample similarity.
• Functional Interpretation: Highlights and prioritizes the gene sets that most significantly contribute to observed clusters, linking clusters to specific biological functions or pathways.
• Comparison to Conventional Methods: Addresses limitations of hierarchical clustering (HC) and consensus clustering (CC) by incorporating gene-set-level signals rather than relying solely on holistic expression profiles.
• Input Requirements: Operates on a list of gene sets and a gene expression data matrix with genes in rows and samples in columns.
• Clustering Integration: Performs clustering on each gene-set-specific subdata and integrates per-gene-set clustering results into a final comprehensive clustering output.
• Validation: Demonstrated improved performance and biologically relevant prioritized gene sets across nine real experimental datasets.

Scientific Applications:

• Cell subpopulation identification: Detects cell subpopulations by grouping samples according to functional gene-set signals in expression data.
• Disease subtype discovery: Identifies disease subtypes and links them to pathway- or function-level gene-set differences.
• Biological interpretation of clusters: Facilitates assigning biological functions or pathways to clusters via prioritized gene sets.

Methodology:

Requires a list of gene sets and a gene expression matrix; partitions the expression matrix into subdata by gene set, performs clustering on each subdata subset, and integrates the per-gene-set clustering results into a final clustering output.