Multiview Robust Graph-Based Clustering for Cancer Subtype Identification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer subtype identification is crucial for personalized diagnosis and therapy.
• Public multi-omics datasets like The Cancer Genome Atlas (TCGA) offer opportunities for comprehensive cancer research.
• Existing methods may be influenced by noise in raw omics data.

Purpose of the Study:

• To propose a novel multi-view robust graph-based clustering (MRGC) method for effective cancer subtype identification.
• To leverage multi-omics data for a more comprehensive understanding of cancer mechanisms.
• To improve the accuracy and robustness of cancer classification.

Main Methods:

• Learns robust latent representations from raw omics data to minimize noise.
• Adaptively learns similarity matrices from these representations.
• Integrates information through a global similarity graph derived from consensus structures.
• Employs a mutually reinforcing iterative process across method components.

Main Results:

• Demonstrates satisfactory clustering performance on generic machine learning and cancer datasets.
• Outperforms several state-of-the-art clustering approaches.
• Validates the practicability of the MRGC method through a case study on hepatocellular carcinoma.

Conclusions:

• The MRGC method effectively identifies cancer subtypes by robustly integrating multi-omics data.
• This approach offers a promising tool for advancing personalized cancer diagnosis and treatment strategies.
• The study highlights the utility of graph-based clustering in complex biological data analysis.