Anomaly Detection with Graph Embedding in Bioinformatics: A Survey.

Area of Science:

• Bioinformatics
• Network Science
• Data Science

Background:

• Biomolecular networks (gene regulatory, protein-protein interaction, drug-molecule) are fundamental to biological processes.
• Abnormal patterns in these networks are linked to diseases, making anomaly detection vital for medical research.
• Graph anomaly detection is a key research area with significant applications in bioinformatics.

Purpose of the Study:

• To review graph embedding-based anomaly detection algorithms in bioinformatics.
• To focus on dynamic graph anomaly detection and its advancements.
• To provide theoretical and methodological support for network anomaly detection in omics data.

Main Methods:

• Review of key literature on graph embedding techniques for biomolecular networks.
• Exploration of how graph embedding learns high-dimensional features and reduces dimensionality.
• Investigation of dynamic graph embedding algorithms for capturing network changes.

Main Results:

• Graph embedding effectively learns features from complex biomolecular networks.
• Dynamic graph embedding algorithms can enhance detection accuracy and efficiency.
• Current methods face challenges that necessitate further research.

Conclusions:

• Graph anomaly detection is critical for disease mechanism elucidation, diagnosis, and drug development.
• Future directions include improving methods for high-throughput omics data analysis.
• This review offers foundational insights for network anomaly detection applications.