VSClust

VSClust performs variance-sensitive clustering of omics data to identify biologically relevant molecular features by incorporating feature-specific variance into cluster assignment.

Key Features:

• Variance-Sensitive Clustering: Accounts for the variance of each feature within a condition instead of averaging measurements to reflect true biological variability.
• Fuzzy-clustering-derived Algorithm: Implements an algorithm derived from fuzzy clustering techniques and is evaluated against fuzzy c-means.
• Integration of Statistical Testing and Pattern Recognition: Unifies statistical testing with pattern recognition to inform cluster formation.
• Scalability to Large Omics Datasets: Applied to artificial and experimental datasets ranging from hundreds to over 80,000 features across 6–20 conditions.
• Handling of Replicated Measurements: Incorporates replicated measurements within conditions to enable robust feature-specific variance assessment and to avoid arbitrary averaging.

Scientific Applications:

• Identification of Functionally Related Features: Groups genes, proteins, and metabolites into clusters to support interpretation of molecular networks and biological processes.
• Multi-omics Analysis: Applied to genomics, transcriptomics, proteomics, and metabolomics datasets for comparative and integrative studies.
• Condition-specific Variation Assessment: Enables robust statistical assessment of feature-specific variation across experimental conditions using replicated measurements.

Methodology:

Uses an algorithm derived from fuzzy clustering techniques that accounts for feature-specific variance, integrates statistical testing with pattern recognition, and avoids averaging measurements within conditions.