vWCluster: Vector-valued optimal transport for network based clustering using multi-omics data in breast cancer.

Area of Science:

• Computational biology
• Network analysis
• Biostatistics

Background:

• Integrating multi-omics data is crucial for understanding complex diseases like breast cancer.
• Existing clustering methods may not effectively handle multi-layer network representations of heterogeneous data.

Purpose of the Study:

• To introduce vector Wasserstein clustering (vWCluster), a novel network-based method for analyzing multi-omics data.
• To apply vWCluster to breast cancer datasets to identify patient subgroups with differential survival and immune microenvironments.

Main Methods:

• Developed vWCluster based on vector-valued Wasserstein distance from optimal mass transport (OMT) theory.
• Applied hierarchical clustering to multi-layer network data representations.
• Utilized Kaplan-Meier analysis and CIBERSORT scores for survival and immune microenvironment analysis.

Main Results:

• vWCluster successfully identified breast cancer subgroups with significantly different survival rates in two large datasets.
• Nine out of 22 CIBERSORT immune cell types were consistently different across identified clusters in both datasets.
• The identified clusters reflect variations in the tumor immune microenvironment.

Conclusions:

• vWCluster effectively integrates heterogeneous multi-omics data in a network structure.
• The method can identify clinically relevant tumor subgroups based on mortality and immune profiles.
• vWCluster offers a powerful approach for precision oncology research.