Subtype-DCC: decoupled contrastive clustering method for cancer subtype identification based on multi-omics data

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Cancer exhibits high heterogeneity, necessitating subtype identification for diagnosis, prognosis, and treatment.
• Multi-omics data availability is increasing due to technological advancements.
• Integrating multi-omics data is essential for comprehensive cancer subtyping.

Purpose of the Study:

• To develop a computational model for effective cancer subtyping using integrated multi-omics data.
• To introduce a novel decoupled contrastive clustering method (Subtype-DCC).

Main Methods:

• Proposed Subtype-DCC, a deep clustering method leveraging contrastive learning.
• Integrated multi-omics data within a single framework using deep neural networks.
• Learned clustering-friendly representations from integrated data.

Main Results:

• Subtype-DCC demonstrated superior performance in identifying cancer subtypes compared to existing methods.
• The model's effectiveness was validated through survival and clinical analyses.
• Successfully integrated complementary information from multi-omics data.

Conclusions:

• Subtype-DCC offers a powerful approach for cancer subtyping.
• The method enhances the potential for personalized cancer diagnosis and treatment.
• Highlights the utility of deep contrastive learning in multi-omics data integration for biomedical applications.