Multi-omics data integration for enhanced cancer subtyping via interactive multi-kernel learning

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Cancer is a complex, heterogeneous disease requiring advanced molecular analysis for effective treatment.
• Multi-omics data integration is crucial for identifying subtypes and enabling personalized medicine.
• Existing methods often fail to capture interactions between different omics data types.

Purpose of the Study:

• To develop a novel method, interactive multi-kernel learning (iMKL), for improved cancer subtype identification.
• To incorporate omics-omics interactions within an unsupervised multi-kernel learning framework.
• To enhance the accuracy and robustness of cancer subtyping using multi-omics data.

Main Methods:

• Developed iMKL, a method integrating omics-omics interactions and heterogeneous data types.
• Utilized a joint Hadamard product strategy to capture higher-order interactive effects.
• Applied iMKL to renal cell carcinoma (RCC) datasets (ccRCC and type II pRCC) with miRNA, mRNA, and DNA methylation data.

Main Results:

• iMKL demonstrated strong robustness and accuracy in identifying patient subtypes via stability analysis.
• Classified both ccRCC and type II pRCC into three distinct subtypes with significant survival differences.
• Identified potential biomarkers associated with adverse patient outcomes.

Conclusions:

• iMKL effectively identifies tumor molecular subtypes strongly associated with clinical features and survival.
• The method offers substantial advancement in cancer subtyping and personalized treatment strategies.
• iMKL provides valuable insights for clinical decision-making in oncology.