PAMOGK: a pathway graph kernel-based multiomics approach for patient clustering.

Area of Science:

• Computational biology and bioinformatics
• Cancer genomics and precision medicine

Background:

• Accurate molecular subtyping of cancer patients is crucial for developing targeted therapies.
• Multiomics data offers detailed insights into tumor biology but requires sophisticated integration methods.

Purpose of the Study:

• To develop a novel computational method, Pathway-based MultiOmic Graph Kernel clustering (PAMOGK), for integrating multiomics data with biological pathway knowledge.
• To improve patient stratification for kidney renal clear cell carcinoma (KIRC) based on molecular profiles.

Main Methods:

• Developed a novel graph kernel to assess patient similarity based on molecular alterations within biological pathways.
• Employed multiview kernel clustering to integrate information from hundreds of pathways and molecular features.
• Applied PAMOGK to multiomics data from KIRC patients.

Main Results:

• PAMOGK identified four distinct KIRC patient clusters with significantly different survival times (P=1.24e-11).
• The discovered subgroups also exhibited differences in tumor stage, grade, and metastatic spread.
• PAMOGK outperformed eight other state-of-the-art multiomics clustering methods in patient stratification.

Conclusions:

• PAMOGK provides a powerful framework for multiomics data integration and patient subgroup discovery.
• The method enhances the identification of clinically relevant molecular subtypes in cancer, specifically KIRC.
• The identified pathways are highly relevant to KIRC pathogenesis, offering potential therapeutic targets.