EMitool: Explainable Multi-Omics Integration for Disease Subtyping.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Personalized medicine requires accurate disease subtyping based on heterogeneity.
• Multi-omics data accumulation drives the need for advanced integration methods.
• Existing methods lack explainability and clear clinical outcome links.

Purpose of the Study:

• To develop EMitool, an explainable multi-omics integration tool.
• To achieve biologically and clinically relevant disease subtyping.
• To link molecular subtypes with clinical outcomes without prior clinical data.

Main Methods:

• Network-based fusion strategy for multi-omics data integration.
• Application to 31 cancer types from The Cancer Genome Atlas (TCGA).
• Comparative analysis against eight state-of-the-art subtyping methods.

Main Results:

• EMitool demonstrated superior subtyping accuracy.
• Identified subtypes significantly associated with survival, stage, and tumor mutational burden.
• Revealed distinct subtypes in kidney renal clear cell carcinoma (KIRC) with differential prognoses and immune profiles.
• Provided contribution scores for omics data types, enhancing interpretability.

Conclusions:

• EMitool enables explainable and clinically relevant multi-omics disease subtyping.
• The tool facilitates biomarker discovery for precision oncology.
• EMitool's subtypes correlate with immune microenvironment and therapeutic responses.