HyperTMO: a trusted multi-omics integration framework based on hypergraph convolutional network for patient...

Area of Science:

• Biomedical informatics
• Computational biology
• Machine learning

Background:

• High-throughput technologies generate detailed multi-omics data for human disease research.
• Machine learning-based multi-omics integration offers comprehensive insights but faces challenges in data representation and integration.

Purpose of the Study:

• To present HyperTMO, a novel framework for Trusted Multi-Omics integration using hypergraph convolutional networks for patient classification.
• To address challenges in representing single-omics data and integrating multi-omics information.

Main Methods:

• HyperTMO constructs hypergraph structures to represent associations within single-omics data.
• Evidence extraction is performed using a hypergraph convolutional network.
• Multi-omics information is integrated at an evidence level.

Main Results:

• HyperTMO outperforms state-of-the-art methods in breast cancer subtype and Alzheimer's disease classification.
• Experiments utilized multi-omics data from TCGA (BRCA) and ROSMAP datasets.
• The framework demonstrates accurate and robust performance.

Conclusions:

• HyperTMO is the first framework to integrate hypergraph structure, evidence theory, and multi-omics integration for patient classification.
• The method shows significant potential for clinical diagnosis applications.
• The framework and datasets are publicly available for further research.