Multi-omics integration with weighted affinity and self-diffusion applied for cancer subtypes identification

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Accurate cancer molecular subtyping is essential for personalized treatment and prognosis.
• Multi-omics data integration offers deeper biological insights but faces computational and data-weighting challenges.
• Existing methods often lack efficiency and robust data type weighting for cancer subtyping.

Purpose of the Study:

• To present MOSD (Multi-omics integration via weighted affinity and Self-Diffusion), an efficient computational method for cancer subtyping.
• To address limitations in computational efficiency and data type weight assignment in current multi-omics integration tools.
• To dissect cancer heterogeneity using multi-omics data for improved clinical applications.

Main Methods:

• Developed MOSD, integrating multi-omics data (gene expression, DNA methylation, miRNA) through weighted affinity and self-diffusion.
• Constructed local scaling affinity for each data type, followed by weighted linear combination for integration.
• Applied self-diffusion to enhance patient similarity for downstream clustering analysis.
• Validated MOSD across ten cancer types.

Main Results:

• MOSD demonstrated significant differences in patient survival across identified molecular subtypes.
• The method showed computational efficiency compared to state-of-the-art integration techniques.
• Identified molecular subtypes exhibited strong biological interpretability.
• MOSD effectively dissected cancer heterogeneity using multi-omics data.

Conclusions:

• MOSD provides an efficient and effective approach for multi-omics data integration in cancer subtyping.
• The method enhances biological interpretability and improves patient survival prediction.
• MOSD offers a valuable tool for understanding cancer heterogeneity and advancing personalized oncology.
• Open-source code is available for reproducibility and further research.