ClustOmics

ClustOmics performs consensus clustering to integrate multi-omic datasets and derive disease subtypes such as cancer classifications.

Key Features:

• Non-Relational Graph Database Integration: Uses a non-relational graph database to manage and integrate diverse omic datasets and clustering outcomes.
• Evidence Accumulation Clustering Strategy: Implements evidence accumulation by computing co-occurrence scores of sample pairs across input clusterings and using those scores as similarity measures to form consensus clusters.
• Flexibility and Robustness: Reconciles clusterings from different origins, sizes, and shapes without requiring a gold-standard metric, providing robustness to heterogeneous input partition quality.
• Application in Cancer Subtyping: Applied to multi-omic disease subtyping using TCGA data from ten different cancer types to produce biologically meaningful clusters.
• Comparison with Existing Tools: Compared with state-of-the-art approaches such as COCA and presented as complementary rather than a direct competitor.

Scientific Applications:

• Cancer Subtyping and Research: Integrates multi-dimensional omic data to identify disease subtypes and investigate cancer heterogeneity.
• Data Integration and Analysis: Reconciles disparate clustering results into consensus clusters to aid interpretation of complex biological datasets.

Methodology:

ClustOmics computes co-occurrence scores for sample pairs across input clusterings (evidence accumulation) as similarity measures to form consensus clusters and stores/integrates omic datasets and clustering outputs in a non-relational graph database.