Effective Subject Representation based on Multi-omics Disease Networks using Graph Embedding.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Evolutionary network modeling is crucial for understanding complex biological systems.
• Multi-omics networks integrate biomolecular interactions (proteins, metabolites) to link to disease phenotypes.
• Accurate subject representation is key for analyzing complex diseases.

Purpose of the Study:

• To introduce an integrative framework for learning node representations from multi-omics networks.
• To combine network-derived node representations with individual biological profiles for enhanced subject representation.
• To improve subject representation accuracy in the context of complex diseases.

Main Methods:

• Developed a novel integrative framework for multi-omics network analysis.
• Employed representation learning techniques for network nodes.
• Integrated learned node embeddings with subject-specific biological data.

Main Results:

• The proposed framework significantly outperforms existing and baseline methods.
• Achieved superior subject representation accuracy, particularly with sparse and structured multi-omics networks.
• Demonstrated the framework's effectiveness on real-world multi-omics data.

Conclusions:

• The integrative framework provides a powerful tool for analyzing multi-omics data.
• Enhanced subject representation aids in understanding complex diseases.
• The method is particularly beneficial for sparse and informative biological networks.