TF-DWGNet: a directed weighted graph neural network with tensor fusion for multi-omics cancer subtype classification

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Multi-omics data integration is crucial for cancer subtype classification.
• Existing methods struggle with data heterogeneity, high dimensionality, and complex dependencies.
• Current graph neural networks lack task-specific directionality and feature interpretability.

Purpose of the Study:

• To develop an effective and interpretable framework for multi-omics data integration in cancer research.
• To address limitations in existing graph neural network approaches for cancer subtype classification.
• To propose TF-DWGNet for improved accuracy and biological insight generation.

Main Methods:

• Proposed TF-DWGNet, a novel Graph Neural Network framework.
• Introduced supervised tree-based Directed Weighted graph construction for each omics modality.
• Implemented a tensor fusion mechanism with low-rank decomposition for efficient interaction capture.

Main Results:

• TF-DWGNet consistently outperformed state-of-the-art baselines on three real-world cancer datasets.
• Demonstrated superior performance across multiple metrics and statistical tests.
• Provided biologically meaningful insights via modality contribution scores and feature importance.

Conclusions:

• TF-DWGNet offers an effective and interpretable solution for multi-omics integration in cancer research.
• The framework successfully models complex data structures and interactions.
• Highlights the potential of directed weighted graphs and tensor fusion for advancing cancer subtype classification.