Bayesian tensor factorization-drive breast cancer subtyping by integrating multi-omics data.

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Breast cancer is a heterogeneous disease requiring precise subtyping for effective treatment.
• Identifying genomic drivers is crucial for personalized breast cancer oncology.
• Current subtyping methods may not fully leverage integrated multi-omics data.

Purpose of the Study:

• To develop and validate a novel computational approach for breast cancer subtyping.
• To integrate multi-omics data (RNA-seq, copy number variation, DNA methylation) for improved subtyping.
• To identify biologically relevant breast cancer subtypes with distinct clinical outcomes.

Main Methods:

• Bayesian tensor factorization (BTF) was employed to integrate multi-omics data from 762 The Cancer Genome Atlas (TCGA) breast cancer patients.
• Consensus clustering was applied to the latent features derived from BTF to identify subtypes.
• Kaplan-Meier (KM) estimators were used to evaluate subtype-specific survival patterns.
• The proposed BTF-based subtyping approach was compared against existing state-of-the-art methods.

Main Results:

• The BTF approach identified 17 optimized latent components, leading to the discovery of six distinct breast cancer subtypes.
• The proposed method demonstrated statistically significant survival differences among the identified subtypes (p < 0.05), outperforming other approaches.
• Identified clusters showed statistically significant distributions, indicating robust subtype discovery.
• The approach effectively utilized publicly available multi-omics data for breast cancer subtyping.

Conclusions:

• The developed Bayesian tensor factorization approach is a promising strategy for breast cancer subtyping.
• Integrating multi-omics data via BTF enables the identification of clinically relevant breast cancer subtypes.
• This method enhances precision oncology by providing a more refined understanding of breast cancer heterogeneity.