An Extensive Assessment of Network Embedding in PPI Network Alignment.

Area of Science:

• Bioinformatics
• Network Science
• Computational Biology

Background:

• Network alignment is crucial for biological network analysis, aiding in discovering conserved pathways and proteins across species.
• Protein-protein interaction (PPI) networks are key structures in biological research.
• Network embedding, modeling nodes as low-dimensional vectors, is an emerging trend.

Purpose of the Study:

• To provide an overview of current protein-protein interaction (PPI) network embedding alignment methods.
• To compare these methods against each other and against classical network alignment algorithms.
• To identify trends and challenges in the field.

Main Methods:

• Literature survey of network embedding alignment methods applied to biological networks.
• Comparative analysis of network embedding and classical network alignment algorithms.
• Examination of algorithm characteristics, including embedding techniques, data sources, and alignment scope (pairwise vs. multi-network).

Main Results:

• Only five network embedding alignment algorithms have been applied to PPI networks, contrasting with numerous classical methods.
• A unified evaluation framework is needed for standardized algorithm comparison.
• Most embedding methods utilize matrix factorization; some leverage external biological data, while others are domain-agnostic.
• Both pairwise and multi-network alignment approaches exist within the surveyed methods.

Conclusions:

• Network embedding for PPI network alignment is an emerging area with potential but requires further development and standardization.
• The field needs more research into unified evaluation metrics and broader application of existing methods.
• Future work should focus on developing robust, domain-agnostic, and scalable network embedding alignment techniques for biological networks.