Supervised graph contrastive learning for cancer subtype identification through multi-omics data integration.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Accurate cancer subtype classification is crucial for effective patient diagnosis, treatment, and prognosis.
• Multi-omics data offers complementary insights but presents challenges due to differing distributions and high dimensionality.
• Integrating diverse omics data for precise cancer classification remains a significant research challenge.

Purpose of the Study:

• To propose a novel method, MCRGCN, for accurate cancer subtype classification by integrating multi-omics data.
• To address the challenges of high dimensionality and varied distributions in omics data.
• To improve the accuracy and clinical relevance of cancer subtype identification.

Main Methods:

• Developed a supervised graph contrast learning method (MCRGCN) for multi-omics data integration.
• Constructed sample networks from multi-omics data and employed residual graph convolution models.
• Utilized supervised contrast loss to ensure consistency across omics features and integrated them for classification.

Main Results:

• MCRGCN demonstrated superior performance in multi-omics data integration compared to existing methods on invasive breast carcinoma (BRCA) and glioblastoma multiforme (GBM) datasets.
• Identified cancer subtypes with significant clinical features through survival analysis.
• Successfully identified potential biomarkers and pathways associated with distinct cancer subtypes.

Conclusions:

• The MCRGCN method effectively integrates multi-omics data for improved cancer subtype classification.
• The identified subtypes possess significant clinical relevance and aid in prognosis.
• The model facilitates the discovery of novel biomarkers and pathways critical for understanding cancer heterogeneity.